@herbertgao/pi-extensions 2026.8.6 → 2026.8.8

package/node_modules/pi-lens/dist/tools/lens-diagnostics.js

@@ -14,6 +14,7 @@ import * as path from "node:path";
 import { Type } from "../clients/deps/typebox.js";
 import { anchorsForDiagnostic, applyDispositions, applyWeakDispositions, getDisposition, } from "../clients/diagnostic-dispositions.js";
 import { applyInlineSuppressions } from "../clients/dispatch/inline-suppressions.js";
+import { gateFindingsByPathFreshness } from "../clients/advisory-provenance.js";
 import { normalizeRuleId } from "../clients/dispatch/rule-id-normalize.js";
 import { applyRulePolicy, rulePolicyMapFromConfig } from "../clients/dispatch/rule-policy.js";
 import { loadPiLensProjectConfig } from "../clients/project-lens-config.js";
@@ -23,18 +24,23 @@ import { getProjectIgnoreMatcher } from "../clients/file-utils.js";
 import { isAtOrAboveHomeDir, normalizeFilePath, } from "../clients/path-utils.js";
 import { getLSPService } from "../clients/lsp/index.js";
 import { primaryServerId } from "../clients/lsp/config.js";
+import { getFullScanWallClockMs } from "../clients/lsp/workspace-sweep-hold.js";
+import { demoteInferredProjectSweepResults } from "../clients/lsp/inferred-project.js";
 import { hashDiagnosticContent, } from "../clients/lsp/diagnostic-binding.js";
 import { loadProjectDiagnosticsDeltaReport, loadProjectDiagnosticsSnapshot, PROJECT_DIAGNOSTICS_CACHE_VERSION, reconcileProjectDiagnosticsSnapshot, } from "../clients/project-diagnostics/cache.js";
-import { warmTriggerFor } from "../clients/project-diagnostics/extractors.js";
-import { fetchFreshProjectDiagnostics } from "../clients/project-diagnostics/fresh-fetch.js";
+import { formatCacheAgeOld, formatNotRunEntry, } from "../clients/project-diagnostics/extractors.js";
+import { ANALYZER_IDS, fetchFreshProjectDiagnostics, } from "../clients/project-diagnostics/fresh-fetch.js";
 import { loadBootstrapClients } from "../clients/bootstrap.js";
 import { generatedSkipNotice, scanTruncationNotice, } from "../clients/lens-engine.js";
 import { scanProjectDiagnostics } from "../clients/project-diagnostics/scanner.js";
-import { getFileDiagnosticSummaries, reconcileScanDiagnostics, reconcileStaleWidgetFiles, } from "../clients/widget-state.js";
+import { getFileDiagnosticSummaries, reconcileCorrelatedScanDiagnostics, reconcileScanDiagnostics, reconcileStaleWidgetDependencyBlockers, reconcileStaleWidgetFiles, widgetDiagnosticUri, } from "../clients/widget-state.js";
+import { logLatency } from "../clients/latency-logger.js";
 import { convertLspDiagnostics } from "../clients/dispatch/utils/lsp-diagnostics.js";
 import { retagAuxiliaryDiagnostics } from "../clients/dispatch/auxiliary-lsp.js";
 import { detectFileRole } from "../clients/file-role.js";
+import { STALE_LINE_MARKER } from "../clients/stale-marker.js";
 import { makeProgressReporter, scanningSummaryLine } from "./scan-progress.js";
+import { demotePastEofDiagnostics, PAST_EOF_STALE_MARKER, resyncDocumentOnPastEof, } from "../clients/diagnostic-line-freshness.js";
 // The widget state exposes the full per-file diagnostic set; this is the tool's
 // own generous display budget per file (independent of the TUI's 12 cap), to
 // keep output bounded on a pathologically broken file.
@@ -45,15 +51,23 @@ const MAX_DIAGNOSTICS_PER_FILE = 50;
 // caller can never believe it checked files it didn't (issue's stated
 // invariant).
 const MAX_PATHS_ENTRIES = 200;
+// #1623: the reason rendered for every heavyweight-analyzer lane (gitleaks,
+// trivy, govulncheck, dead-code, knip, jscpd, madge, opengrep, test-runner)
+// when mode=full is called without refreshRunners=cheap/all/cached — the
+// "quick mode" gate that skips the expensive fresh-fetch entirely (see
+// `formatFullMode`'s `analyzersPromise`). One shared string so it renders
+// identically everywhere it's used.
+const NOT_REQUESTED_REASON = "refreshRunners not requested this call (quick mode) — pass refreshRunners=cheap/all/cached to lens_diagnostics mode=full to run it";
 // Wall-clock ceiling for the whole mode=full scan. Even with per-file budgets
 // and abort-signal honoring, a pathological state (a language server hanging on
 // spawn/initialize across many files) could otherwise stall the tool
 // indefinitely — an unattended session was observed hung for ~8h. This hard cap
 // aborts the scan so it always returns (partial) rather than never. Env-tunable.
-const FULL_SCAN_WALL_CLOCK_MS = (() => {
-    const raw = Number(process.env.PI_LENS_LENS_DIAGNOSTICS_FULL_TIMEOUT_MS);
-    return Number.isFinite(raw) && raw > 0 ? raw : 300_000; // 5 min default
-})();
+// #1618: single source with `clients/runtime-turn.ts`'s idle-reset base delay
+// (see `getBaseLspIdleResetMs`), which derives itself from this value so the
+// two constants can no longer drift back into the idle-reset-fires-mid-sweep
+// relationship that caused #1618 in the first place.
+const FULL_SCAN_WALL_CLOCK_MS = getFullScanWallClockMs();
 // Binding verification is a fallback for result producers that supplied a
 // content hash without an already-computed disk verdict. Keep it one-file-at-a-
 // time and yield periodically: a full scan can contain thousands of results,
@@ -174,6 +188,17 @@ captureLspStatusRepaint, getRuntime) {
                     return `lens_diagnostics delta — clean${coldSuffix}${failedSuffix}`;
                 return `lens_diagnostics delta — ${aw} actionable · ${cq} quality · ${pd} project${coldSuffix}${failedSuffix}`;
             }
+            // #1799: `semantic === "blocking"` iff `severity === "error"` holds for
+            // RAW classification, but not for the rendered counts — a
+            // dependency-drift demotion (#1631, widget-state.ts `isBlocking` vs
+            // `countDiagnostics`) revokes an error's blocking authority (stale)
+            // while the finding itself stays real evidence, so it lands in
+            // `totalErrors` with `totalBlocking` unchanged. That is the one case
+            // where the two totals genuinely disagree, and it must still render —
+            // otherwise a file with drift-demoted errors reads as "clean". What
+            // must NOT happen is printing both terms for the SAME findings (the
+            // actual #1799 bug), so `errors` renders only when `blocking === 0`,
+            // mirroring the per-file row's own guard below.
             const b = details?.totalBlocking ?? 0;
             const e = details?.totalErrors ?? 0;
             const w = details?.totalWarnings ?? 0;
@@ -181,7 +206,11 @@ captureLspStatusRepaint, getRuntime) {
             if (b + e + w === 0) {
                 return `lens_diagnostics ${mode} — clean (${files} files)${coldSuffix}${failedSuffix}`;
             }
-            return `lens_diagnostics ${mode} — ${b} blocking · ${e} errors · ${w} warnings (${files} files)${coldSuffix}${failedSuffix}`;
+            const parts = [`${b} blocking`];
+            if (b === 0 && e > 0)
+                parts.push(`${e} errors`);
+            parts.push(`${w} warnings`);
+            return `lens_diagnostics ${mode} — ${parts.join(" · ")} (${files} files)${coldSuffix}${failedSuffix}`;
         }),
         parameters: Type.Object({
             mode: Type.Optional(Type.String({
@@ -271,7 +300,25 @@ captureLspStatusRepaint, getRuntime) {
             await flushPending();
             const staleDropped = await reconcileStaleWidgetFiles();
             if (mode === "all") {
-                return formatAllMode(cwd, severity, undefined, undefined, staleDropped, pathsScope);
+                // #1631 dependency-axis gate for the mode=all store. `reconcileStaleWidgetFiles`
+                // (above) only sees the diagnosed file's OWN mtime; this demotes blocking
+                // findings whose forward imports drifted out-of-band. Paired reconciles, one
+                // read site — the same gate that covers the turn-end inline-blocker store.
+                const dependencyGateStart = Date.now();
+                const { demoted: dependencyDemoted, truncatedImports } = await reconcileStaleWidgetDependencyBlockers(cwd, getRuntime?.()?.turnIndex);
+                logLatency({
+                    type: "phase",
+                    toolName: "lens_diagnostics",
+                    filePath: cwd,
+                    phase: "blocker_freshness_widget_gate",
+                    durationMs: Date.now() - dependencyGateStart,
+                    metadata: {
+                        demoted: dependencyDemoted,
+                        truncatedImports,
+                        surface: "mode=all",
+                    },
+                });
+                return formatAllMode(cwd, severity, undefined, undefined, staleDropped, pathsScope, dependencyDemoted);
             }
             if (mode === "full") {
                 // Fold a hard wall-clock ceiling into the abort signal so the scan
@@ -301,16 +348,48 @@ captureLspStatusRepaint, getRuntime) {
     };
 }
 // ── delta mode ────────────────────────────────────────────────────────────────
-function formatProjectDeltaDiagnostic(diagnostic) {
-    const marker = diagnostic.semantic === "blocking" || diagnostic.severity === "error"
-        ? "🔴"
-        : "ℹ";
+function formatProjectDeltaDiagnostic(diagnostic, stale) {
+    // #1944: a demoted row loses the authority marker along with its
+    // coordinate. It kept the 🔴 while its line was replaced by the stale
+    // marker, which is the same "changed the channel, not the body" defect the
+    // turn-end advisory carried — `formatFullMode` (this file, the `d.stale`
+    // arm of its marker) already drops the marker for exactly this reason.
+    const marker = stale
+        ? "○"
+        : diagnostic.semantic === "blocking" || diagnostic.severity === "error"
+            ? "🔴"
+            : "ℹ";
     const rule = diagnostic.rule ?? diagnostic.code ?? diagnostic.runner;
-    return `  ${marker} L${diagnostic.line ?? "?"}  ${rule}  ${diagnostic.message}`;
+    const where = stale ? STALE_LINE_MARKER : `L${diagnostic.line ?? "?"}`;
+    return `  ${marker} ${where}  ${rule}  ${diagnostic.message}`;
 }
+/**
+ * #1634 review round R3: `appendProjectDiagnosticsDeltaLines` re-serves the
+ * persisted project-diagnostics DELTA report verbatim — the third of
+ * `mode=delta`'s three arms (the other two, actionable/quality warnings, were
+ * gated in the F3 pass above), and it carried the identical unfixed shape:
+ * cited `file:line` findings from a report with its own `generatedAt`, no
+ * freshness check. Same gate, same policy as `applyDeltaFreshnessGate`:
+ * missing file -> DROP, edited-since -> DEMOTE (loses its line, keeps
+ * rule/message), live -> unchanged. See `clients/finding-delivery-gate.ts`
+ * surface `lens-diagnostics:mode-delta`.
+ */
 function appendProjectDiagnosticsDeltaLines(lines, cwd, report, severity, includeFile) {
-    const diagnostics = (report?.diagnostics ?? []).filter((diagnostic) => includeFile(diagnostic.filePath) &&
-        matchesSeverity(projectDiagnosticToWidget(diagnostic), severity));
+    const scoped = (report?.diagnostics ?? []).filter((diagnostic) => includeFile(diagnostic.filePath) &&
+        matchesSeverity(projectDiagnosticToWidget(diagnostic, diagnostic.filePath), severity));
+    const gated = gateFindingsByPathFreshness({
+        store: "lens-diagnostics-delta-project",
+        findings: scoped,
+        cwd,
+        scannedAt: report?.generatedAt,
+        citedPath: (d) => d.filePath,
+        onMissing: "drop",
+    });
+    const staleSet = new Set(gated.stale);
+    // Concatenating live-then-stale reorders a file's demoted rows to the end
+    // of its bucket below (rather than each diagnostic's original report
+    // order) — cosmetic only, findings are neither dropped nor duplicated.
+    const diagnostics = [...gated.live, ...gated.stale];
     const byFile = new Map();
     for (const diagnostic of diagnostics) {
         const filePath = path.resolve(diagnostic.filePath);
@@ -323,7 +402,7 @@ function appendProjectDiagnosticsDeltaLines(lines, cwd, report, severity, includ
         if (!lines.includes(rel))
             lines.push(rel);
         for (const diagnostic of fileDiagnostics) {
-            lines.push(formatProjectDeltaDiagnostic(diagnostic));
+            lines.push(formatProjectDeltaDiagnostic(diagnostic, staleSet.has(diagnostic)));
         }
     }
     return diagnostics.length;
@@ -342,7 +421,7 @@ function appendProjectDiagnosticsDeltaLines(lines, cwd, report, severity, includ
 function filterDeltaReportDispositions(report, cwd, policyMap) {
     if (!report?.diagnostics.length)
         return report;
-    const kept = report.diagnostics.filter((d) => applyWeakDispositions([projectDiagnosticToWidget(d)], cwd, d.filePath)
+    const kept = report.diagnostics.filter((d) => applyWeakDispositions([projectDiagnosticToWidget(d, d.filePath)], cwd, d.filePath)
         .length === 1);
     const policyKept = applyRulePolicy(kept, policyMap);
     if (policyKept.length === report.diagnostics.length)
@@ -360,6 +439,58 @@ function filterDeltaReportDispositions(report, cwd, policyMap) {
 function loadProjectRulePolicyMap(cwd) {
     return rulePolicyMapFromConfig(loadPiLensProjectConfig(cwd).rules);
 }
+/**
+ * #1634 review round: `formatDeltaMode` re-serves the `actionable-warnings`/
+ * `code-quality-warnings` caches verbatim — each cited `file:line`, no
+ * freshness check — the SAME shape #1622 fixed for gitleaks/trivy-secrets,
+ * unfixed here because `mode=delta` is the tool's DEFAULT and easy to miss
+ * in a per-store sweep. Both reports carry a report-level `generatedAt`; a
+ * warning is gated against it exactly like a scanner finding against
+ * `scannedAt`: missing file → DROP (nothing to fix in a file that's gone),
+ * edited-since → DEMOTE (loses its line, keeps its rule/message, marked
+ * `STALE_LINE_MARKER`), live → unchanged. See
+ * `clients/finding-delivery-gate.ts` surface `lens-diagnostics:mode-delta`.
+ */
+function applyDeltaFreshnessGate(files, cwd, generatedAt) {
+    if (!generatedAt)
+        return files;
+    const flat = [];
+    for (const file of files) {
+        for (const warning of file.warnings)
+            flat.push({ filePath: file.filePath, warning });
+    }
+    if (flat.length === 0)
+        return files;
+    const gated = gateFindingsByPathFreshness({
+        store: "lens-diagnostics-delta",
+        findings: flat,
+        cwd,
+        scannedAt: generatedAt,
+        citedPath: (f) => f.filePath,
+        onMissing: "drop",
+    });
+    // Two passes (live, then stale) reorder a file's demoted rows to the end
+    // of its warnings array rather than the original report order — cosmetic
+    // only, nothing is dropped or duplicated.
+    const byFile = new Map();
+    for (const f of gated.live) {
+        const arr = byFile.get(f.filePath) ?? [];
+        arr.push(f.warning);
+        byFile.set(f.filePath, arr);
+    }
+    for (const f of gated.stale) {
+        const arr = byFile.get(f.filePath) ?? [];
+        arr.push({ ...f.warning, stale: true, line: undefined });
+        byFile.set(f.filePath, arr);
+    }
+    return files
+        .map((file) => ({
+        filePath: file.filePath,
+        warnings: byFile.get(file.filePath) ?? [],
+    }))
+        .filter((file) => file.warnings.length > 0);
+}
+// @delivery-surface: lens-diagnostics:mode-delta
 function formatDeltaMode(cacheManager, cwd, severity, pathsScope) {
     const actionableEntry = cacheManager.readCache("actionable-warnings", cwd);
     const qualityEntry = cacheManager.readCache("code-quality-warnings", cwd);
@@ -387,8 +518,8 @@ function formatDeltaMode(cacheManager, cwd, severity, pathsScope) {
         warnings: applyRulePolicy(applyWeakDispositions(file.warnings ?? [], cwd, file.filePath), policyMap),
     }))
         .filter((file) => file.warnings.length > 0);
-    const actionableFiles = visibleWarningFiles(actionable?.files);
-    const qualityFiles = visibleWarningFiles(quality?.files);
+    const actionableFiles = applyDeltaFreshnessGate(visibleWarningFiles(actionable?.files), cwd, actionable?.generatedAt);
+    const qualityFiles = applyDeltaFreshnessGate(visibleWarningFiles(quality?.files), cwd, quality?.generatedAt);
     const lines = [];
     // Fixable warnings from actionable-warnings
     if (severity !== "error") {
@@ -396,7 +527,8 @@ function formatDeltaMode(cacheManager, cwd, severity, pathsScope) {
             const rel = path.relative(cwd, file.filePath);
             lines.push(`${rel}`);
             for (const w of file.warnings) {
-                lines.push(`  ⚠ L${w.line ?? "?"}  ${w.rule ?? w.code ?? w.tool}  ${w.message}`);
+                const where = w.stale ? STALE_LINE_MARKER : `L${w.line ?? "?"}`;
+                lines.push(`  ⚠ ${where}  ${w.rule ?? w.code ?? w.tool}  ${w.message}`);
             }
         }
     }
@@ -407,7 +539,8 @@ function formatDeltaMode(cacheManager, cwd, severity, pathsScope) {
             if (!lines.includes(rel))
                 lines.push(rel);
             for (const w of file.warnings) {
-                lines.push(`  ℹ L${w.line ?? "?"}  ${w.rule ?? w.code ?? w.tool}  ${w.message}`);
+                const where = w.stale ? STALE_LINE_MARKER : `L${w.line ?? "?"}`;
+                lines.push(`  ℹ ${where}  ${w.rule ?? w.code ?? w.tool}  ${w.message}`);
             }
         }
     }
@@ -568,8 +701,12 @@ function applyProjectRulePolicy(value, policyMap) {
         return undefined;
     return { ...value, diagnostics: applyRulePolicy(value.diagnostics, policyMap) };
 }
-/** A diagnostic counts as error-like when it blocks or has error severity. */
+/** A diagnostic counts as error-like when it blocks or has error severity.
+ * A `stale` (#1641 past-EOF) entry is excluded — its cited line no longer
+ * exists in the current file, so it can no longer count toward a hard stop. */
 function isErrorLike(d) {
+    if (d.stale)
+        return false;
     return d.semantic === "blocking" || d.severity === "error";
 }
 function matchesSeverity(d, severity) {
@@ -620,7 +757,7 @@ function lspDiagnosticToWidget(diagnostic) {
         tool: "lsp",
     };
 }
-function projectDiagnosticToWidget(diagnostic) {
+function projectDiagnosticToWidget(diagnostic, filePath) {
     return {
         severity: diagnostic.severity,
         semantic: diagnostic.semantic,
@@ -629,6 +766,7 @@ function projectDiagnosticToWidget(diagnostic) {
         col: diagnostic.column,
         rule: diagnostic.rule ?? diagnostic.code,
         tool: diagnostic.runner || diagnostic.tool,
+        uri: widgetDiagnosticUri(filePath, diagnostic.line, diagnostic.column),
     };
 }
 // The ast-grep LSP keys its diagnostics `rule = "ast-grep:<id>"` (source:code,
@@ -650,11 +788,23 @@ function summarizeDiagnostics(filePath, diagnostics, hasFinalSnapshot) {
     let errors = 0;
     let warnings = 0;
     for (const diagnostic of diagnostics) {
+        // #1641: a demoted past-EOF entry keeps its severity for display but
+        // drops out of the tallies — same reasoning as widget-state's `isBlocking`.
+        if (diagnostic.stale)
+            continue;
         if (diagnostic.semantic === "blocking")
             blocking++;
         if (diagnostic.severity === "error")
             errors++;
-        else if (diagnostic.severity === "warning")
+        // #1777: `hint` and `info` tally alongside `warning`, matching
+        // `countDiagnostics` in `clients/widget-state.ts` — mode=all reads that
+        // tally and mode=full reads this one, so the two must agree. An exact
+        // `=== "warning"` test scored a hint-only file 0/0/0, the `withIssues`
+        // filter below then dropped it, and mode=full rendered "No issues" for a
+        // file whose own `details` still carried the diagnostic.
+        else if (diagnostic.severity === "warning" ||
+            diagnostic.severity === "hint" ||
+            diagnostic.severity === "info")
             warnings++;
     }
     return {
@@ -666,6 +816,85 @@ function summarizeDiagnostics(filePath, diagnostics, hasFinalSnapshot) {
         diagnostics,
     };
 }
+// #1618: sentence form used when exactly one reason accounts for every
+// unconfirmed file — mirrors the pre-#1618 two-case ternary's phrasing so
+// existing single-reason callers see unchanged text.
+const UNCONFIRMED_REASON_SENTENCE = {
+    budget: "check didn't complete within budget",
+    inconclusive: "check was inconclusive (notify-write or diagnostics wait timed out)",
+    coverage_gap: "an auxiliary scanner coverage gap",
+    service_destroyed: "the LSP service was reset mid-sweep",
+    error: "check errored",
+    // #1618 review round 2: the LSP layer considered this touch CONFIRMED —
+    // the file's content changed under it (stale binding), which is a
+    // completely different failure from a timeout. Must never collapse into
+    // "within budget", the exact string this PR exists to stop misrendering.
+    binding_mismatch: "the file changed on disk since this result was computed",
+};
+// Short per-reason label used only when MORE than one reason is present, so
+// the note can still say "N timed out, M errored" instead of collapsing to
+// one sentence that can't carry two counts.
+const UNCONFIRMED_REASON_COUNT_LABEL = {
+    budget: "timed out",
+    inconclusive: "inconclusive",
+    coverage_gap: "coverage gap",
+    service_destroyed: "service reset mid-sweep",
+    error: "errored",
+    binding_mismatch: "stale binding",
+};
+/**
+ * #1618: a result predating the `unconfirmedReason` field (a legacy/test
+ * double, or `LSPWorkspaceDiagnosticResult`'s own pre-#1618 shape) is
+ * classified from its OTHER fields exactly the way the dead
+ * `unconfirmedErrored = length - unconfirmedTimedOut` subtraction meant to,
+ * before that subtraction was structurally always 0 (the sweep's catch
+ * block sets both `error` and `timedOut`, so `unconfirmedErrored` never had
+ * a way to become nonzero). An explicit `result.error` now wins even when
+ * `unconfirmedReason` is absent, so an errored file is never rendered as
+ * "within budget".
+ *
+ * #1618 review round 2: `mismatched` MUST be checked first. A binding
+ * mismatch is discovered HERE, after the sweep already returned the result
+ * as confirmed (`!timedOut`, no `.error`, no `.unconfirmedReason`) — the
+ * `?? "budget"` fallback below would otherwise silently claim a
+ * stale-binding file as "within budget", the exact string this PR exists to
+ * stop misrendering.
+ */
+function classifyUnconfirmedReason(result, mismatched) {
+    if (mismatched.has(result))
+        return "binding_mismatch";
+    return result.unconfirmedReason ?? (result.error ? "error" : "budget");
+}
+/**
+ * #1618: replaces the `unconfirmedTimedOut`/`unconfirmedErrored` dead
+ * subtraction with a real per-reason tally read off each result's own
+ * `unconfirmedReason` (falling back to `classifyUnconfirmedReason` for a
+ * legacy result that predates the field). `mismatched` is threaded through
+ * so a stale-binding file is never counted under a fallback reason it
+ * doesn't have (review round 2).
+ */
+function tallyUnconfirmedReasons(results, mismatched) {
+    const tally = new Map();
+    for (const result of results) {
+        const reason = classifyUnconfirmedReason(result, mismatched);
+        tally.set(reason, (tally.get(reason) ?? 0) + 1);
+    }
+    return tally;
+}
+/** Render `tallyUnconfirmedReasons`' output as the clause inside the
+ *  "unconfirmed (...)" note — a single sentence when one reason accounts for
+ *  everything, else a comma-joined "N label" breakdown. */
+function formatUnconfirmedReasonClause(tally) {
+    const entries = [...tally.entries()].filter(([, count]) => count > 0);
+    if (entries.length === 0)
+        return "";
+    if (entries.length === 1) {
+        return UNCONFIRMED_REASON_SENTENCE[entries[0][0]];
+    }
+    return entries
+        .map(([reason, count]) => `${count} ${UNCONFIRMED_REASON_COUNT_LABEL[reason]}`)
+        .join(", ");
+}
 /**
  * #646: break the confirmed/unconfirmed LSP-sweep tally down PER primary
  * server id (typescript, pyright, marksman, ...) instead of one flat
@@ -770,10 +999,10 @@ function mergeDiagnosticsWithWidgetSummaries(widgetSummaries, lspResults, projec
         }
     }
     for (const diagnostic of projectSnapshot?.diagnostics ?? []) {
-        addDiagnostic(path.resolve(diagnostic.filePath), projectDiagnosticToWidget(diagnostic));
+        addDiagnostic(path.resolve(diagnostic.filePath), projectDiagnosticToWidget(diagnostic, diagnostic.filePath));
     }
     for (const diagnostic of projectDelta?.diagnostics ?? []) {
-        addDiagnostic(path.resolve(diagnostic.filePath), projectDiagnosticToWidget(diagnostic));
+        addDiagnostic(path.resolve(diagnostic.filePath), projectDiagnosticToWidget(diagnostic, diagnostic.filePath));
     }
     return [...byFile.values()];
 }
@@ -896,6 +1125,7 @@ async function getProjectDiagnosticsSnapshotForFullMode(cwd, options) {
     }
     return undefined;
 }
+// @delivery-surface: lens-diagnostics:mode-full
 async function formatFullMode(cwd, severity, lspService, cacheManager, options = {}) {
     const runWorkspaceDiagnostics = lspService.runWorkspaceDiagnostics;
     if (typeof runWorkspaceDiagnostics !== "function") {
@@ -935,12 +1165,22 @@ async function formatFullMode(cwd, severity, lspService, cacheManager, options =
     // could complete. Building its promise here, alongside the sweep and cheap
     // scan, makes it genuinely concurrent — all three phases now race the SAME
     // signal from the same starting point instead of stacking.
-    const analyzersPromise = shouldIncludeProjectRunners(options.refreshRunners)
+    const projectRunnersRequested = shouldIncludeProjectRunners(options.refreshRunners);
+    const analyzersPromise = projectRunnersRequested
         ? loadBootstrapClients().then((clients) => fetchFreshProjectDiagnostics(cacheManager, cwd, clients, signal, { runtime: options.runtime }))
         : Promise.resolve({
             diagnostics: [],
             runners: [],
-            cold: [],
+            // #1623: every heavyweight analyzer is ELIGIBLE for this project but
+            // this call never asked for it (refreshRunners wasn't cheap/all/
+            // cached) — the expensive fetch below deliberately never runs in
+            // that case (see the comment above), but rendering total silence
+            // here reads exactly as "ran clean", the false-empty #1623 exists to
+            // close. `ANALYZER_IDS` (fresh-fetch.ts) is the single source of
+            // truth for which lanes this covers — reuse it rather than
+            // hand-listing ids here.
+            cold: [...ANALYZER_IDS],
+            coldReasons: Object.fromEntries(ANALYZER_IDS.map((id) => [id, NOT_REQUESTED_REASON])),
             failed: [],
             timings: {},
         });
@@ -960,7 +1200,14 @@ async function formatFullMode(cwd, severity, lspService, cacheManager, options =
         analyzersPromise,
     ]);
     const aborted = signal?.aborted ?? false;
-    const lspResults = rawLspResults.filter((result) => includeFile(result.filePath));
+    // #1640: before ANY consumer sees them — the footer reconcile, the widget
+    // merge, the rendered counts — demote TypeScript errors on files tsserver
+    // checked in its INFERRED project. Applied once, here, so a single seam
+    // governs the whole sweep instead of each renderer learning the rule.
+    const lspResults = await demoteInferredProjectSweepResults(rawLspResults.filter((result) => includeFile(result.filePath)), cwd, lspService, 
+    // #1645 review F1: the sweep that produced these results is signal-bounded,
+    // so the probe loop that post-processes them must be too.
+    signal);
     // A result bound to a different document must not replace the current
     // widget state, even when it contains real diagnostics. Pull results may carry
     // a content hash without the push binding verdict, so verify that hash against
@@ -981,13 +1228,15 @@ async function formatFullMode(cwd, severity, lspService, cacheManager, options =
     const confirmedLspResults = lspResults.filter((result) => !result.timedOut &&
         !result.error &&
         !mismatchedLspResults.has(result));
+    const fullyCoveredLspResults = confirmedLspResults.filter((result) => (result.unconfirmedServerIds?.length ?? 0) === 0);
+    const partiallyCoveredLspResults = confirmedLspResults.filter((result) => (result.unconfirmedServerIds?.length ?? 0) > 0);
     const unconfirmedLspResults = lspResults.filter((result) => result.timedOut ||
         result.error ||
         mismatchedLspResults.has(result));
     // #571: reconcile this scan's fresh, CONFIRMED per-file results into the
     // footer cache. A footer write is never allowed to fail the tool call, so
     // any unexpected throw is swallowed.
-    for (const result of confirmedLspResults) {
+    for (const result of fullyCoveredLspResults) {
         try {
             // #692: provenance label ONLY — must never affect `rule`/identity (see
             // `ConvertLspDiagnosticsOptions.scanOrigin`'s doc comment).
@@ -1053,6 +1302,22 @@ async function formatFullMode(cwd, severity, lspService, cacheManager, options =
     // legitimately show diagnostics from widgetSummaries/project-runner state
     // below if those independently have entries for it.
     const summaries = await applyInlineSuppressionsToSummaries(mergeDiagnosticsWithWidgetSummaries(getFileDiagnosticSummaries().filter((summary) => includeFile(summary.filePath)), confirmedLspResults, projectSnapshot, projectDelta), cwd, policyMap);
+    // #1888: the full-mode summary above is the first seam where every
+    // producing lane is correlated. The earlier footer loop intentionally writes
+    // only CONFIRMED LSP results, so an ast-grep backpressure failure left
+    // project-scan (`ast-grep-napi`) findings visible to lens_diagnostics but
+    // absent from the widget. Commit this same post-policy, post-suppression set
+    // to the widget so its human-facing count cannot silently become single-lane.
+    const parsedProjectScanObservedAt = projectSnapshot?.scannedAt
+        ? Date.parse(projectSnapshot.scannedAt)
+        : undefined;
+    const projectScanObservedAt = parsedProjectScanObservedAt !== undefined &&
+        Number.isFinite(parsedProjectScanObservedAt)
+        ? parsedProjectScanObservedAt
+        : undefined;
+    for (const summary of summaries) {
+        reconcileCorrelatedScanDiagnostics(summary.filePath, summary.diagnostics, nextWriteIndex?.(), projectScanObservedAt);
+    }
     const result = formatAllMode(cwd, severity, summaries, {
         mode: "full",
         lspFilesChecked: rawLspResults.length,
@@ -1083,8 +1348,7 @@ async function formatFullMode(cwd, severity, lspService, cacheManager, options =
     // as "0 diagnostics" — say explicitly which files the LSP sweep could not
     // confirm and why, distinguishing a hard error from a soft timeout the
     // same way #570 does upstream.
-    const unconfirmedTimedOut = unconfirmedLspResults.filter((result) => result.timedOut).length;
-    const unconfirmedErrored = unconfirmedLspResults.length - unconfirmedTimedOut;
+    const unconfirmedByReason = tallyUnconfirmedReasons(unconfirmedLspResults, mismatchedLspResults);
     // #646: per-primary-server breakdown of the same confirmed/unconfirmed
     // tally — lets an agent see AT A GLANCE which server is responsible for
     // any unconfirmed files (e.g. "marksman: 0/34") instead of having to
@@ -1099,15 +1363,21 @@ async function formatFullMode(cwd, severity, lspService, cacheManager, options =
         : "";
     const unconfirmedLspNote = unconfirmedLspResults.length > 0
         ? `\n\n⚠ LSP sweep: ${confirmedLspResults.length} file(s) confirmed via LSP, ` +
-            `${unconfirmedLspResults.length} unconfirmed (${unconfirmedTimedOut > 0 && unconfirmedErrored > 0
-                ? `${unconfirmedTimedOut} timed out, ${unconfirmedErrored} errored`
-                : unconfirmedTimedOut > 0
-                    ? "check didn't complete within budget"
-                    : "check errored"})${serverBreakdownClause} — NOT the same as 0 diagnostics for: ${unconfirmedLspResults
+            `${unconfirmedLspResults.length} unconfirmed (${formatUnconfirmedReasonClause(unconfirmedByReason)})${serverBreakdownClause} — NOT the same as 0 diagnostics for: ${unconfirmedLspResults
                 .slice(0, 20)
                 .map((result) => result.filePath)
                 .join(", ")}${unconfirmedLspResults.length > 20 ? ", …" : ""}. These files' LSP contribution is excluded from this result; re-run mode=full to retry them (they may still show findings above from cached/project-runner state).`
         : "";
+    // Code-unit comparator (#1883): this list ships as the
+    // `unconfirmedLspServerIds` structured field and as the lane names in the
+    // agent-visible coverage note, so its order must be deterministic across
+    // locales — localeCompare is deliberately avoided.
+    const uncoveredScannerIds = [
+        ...new Set(partiallyCoveredLspResults.flatMap((result) => result.unconfirmedServerIds ?? [])),
+    ].sort((a, b) => Number(a > b) - Number(a < b));
+    const auxiliaryCoverageNote = uncoveredScannerIds.length > 0
+        ? `\n\n⚠ Auxiliary coverage incomplete: ${uncoveredScannerIds.join(", ")} did not answer for ${partiallyCoveredLspResults.length} file(s). Findings from answering servers are included; this is not a clean verdict for the named scanner lane(s).`
+        : "";
     // #646: primary-vs-auxiliary split of the raw LSP-sweep findings (before
     // they're merged into the widget-state summaries below), mirroring
     // `lsp_diagnostics`' "Primary findings"/"Auxiliary findings" separation —
@@ -1139,13 +1409,31 @@ async function formatFullMode(cwd, severity, lspService, cacheManager, options =
     // analyzers before anything spawns — rendering that as the per-analyzer
     // "not applicable / unavailable" list would send the caller chasing seven
     // wrong reasons. One note with the real one instead.
+    // #1623: prefer the SPECIFIC reason captured at each gate
+    // (`extracted.coldReasons`) over a generic guess — `formatNotRunEntry`
+    // falls back to `warmTriggerFor` only for an id this call's `coldReasons`
+    // doesn't cover (e.g. a caller-supplied `cold` list from an older cache
+    // shape). Single formatter (extractors.ts) so this note's wording can't
+    // drift from any other caller that renders the same `cold` list.
+    // #1623 fix-round F5: the "not requested" (quick-mode) batch shares ONE
+    // reason (`NOT_REQUESTED_REASON`) across every id — `formatNotRunEntry`
+    // repeating that same ~130-char sentence per id makes the note nearly
+    // unreadable, and the "not applicable / unavailable this run" header
+    // below is a dishonest label for it: these lanes ARE applicable and
+    // available, this call just never asked for them. Render that batch as
+    // its own compact note — bare ids, the shared reason stated once — and
+    // keep the detailed per-id format for genuinely cold lanes, where each
+    // id's reason really does differ (not a git repo vs binary unavailable
+    // vs retry cooldown, ...).
     const coldNote = extracted.unsafeRoot
         ? `\n\nheavyweight analyzers skipped: the working directory resolves at or above the home directory, so a fresh knip/jscpd/madge/gitleaks/govulncheck/trivy/dead-code scan would walk every unrelated tree under it. Re-run from inside a project directory. Absence of their findings is NOT a clean verdict.`
-        : genuinelyColdIds.length > 0
-            ? `\n\ncold (not applicable / unavailable this run): ${genuinelyColdIds
-                .map((id) => `${id} — ${warmTriggerFor(id)}`)
-                .join(", ")}. These analyzers have not contributed to this result — absence of their findings is NOT a clean verdict.`
-            : "";
+        : !projectRunnersRequested && genuinelyColdIds.length > 0
+            ? `\n\nnot run this call (quick mode): ${genuinelyColdIds.join(", ")}. ${NOT_REQUESTED_REASON}. Absence of their findings is NOT a clean verdict.`
+            : genuinelyColdIds.length > 0
+                ? `\n\ncold (not applicable / unavailable this run): ${genuinelyColdIds
+                    .map((id) => formatNotRunEntry(id, extracted.coldReasons))
+                    .join(", ")}. These analyzers have not contributed to this result — absence of their findings is NOT a clean verdict.`
+                : "";
     // #1004: unlike every other analyzer above (knip/jscpd/madge/gitleaks/
     // govulncheck/opengrep/trivy/dead-code all run a FRESH whole-project scan
     // per the fresh-fetch.ts header, so "clean" there really does mean "the
@@ -1170,6 +1458,21 @@ async function formatFullMode(cwd, severity, lspService, cacheManager, options =
             .map(({ id, summary }) => `${id} — ${summary}`)
             .join(", ")}. These analyzers were not cached and will be retried; absence of their findings is NOT a clean verdict.`
         : "";
+    // #1617: the #1616 suppressed-bucket rule applied to mode=full — a finding
+    // an agent/user marked false-positive/won't-fix now drops out of
+    // `diagnostics` (fresh-fetch.ts's `record()`), but that drop must stay
+    // visible as a count rather than reading as "nothing was wrong here".
+    // Review-round F4 (#1625): per-lane attribution — "gitleaks 2, knip 1" says
+    // WHICH analyzer's marks are doing the suppressing, not just a bare total.
+    // A lane that's 100% suppressed still won't appear in `runners`/`cold` as
+    // distinct from "ran clean" — that gap is #1623's lane-status territory,
+    // not fixed here.
+    const dispositionSuppressedByLaneText = Object.entries(extracted.dispositionSuppressedByLane ?? {})
+        .map(([id, count]) => `${id} ${count}`)
+        .join(", ");
+    const dispositionSuppressedNote = (extracted.dispositionSuppressed ?? 0) > 0
+        ? `\n\nsuppressed by disposition: ${extracted.dispositionSuppressed} finding(s) dropped from this result because they're marked false-positive or won't-fix (${dispositionSuppressedByLaneText}). Not a clean verdict for those locations — they were found, then intentionally hidden.`
+        : "";
     // #747/#250: the cheap project-diagnostics scan (scanProjectDiagnostics) and
     // the LSP workspace sweep (collectWorkspaceDiagnosticFiles) both refuse to
     // WALK from a cwd at/above $HOME — from there, walking would enumerate every
@@ -1199,6 +1502,28 @@ async function formatFullMode(cwd, severity, lspService, cacheManager, options =
     const generatedSkipNote = generatedSkipNoticeText
         ? `\n\n${generatedSkipNoticeText}`
         : "";
+    // #1623: the cheap in-process project scan (tree-sitter/fact-rules/
+    // ast-grep) is a THIRD lane `coldNote` never covers (it isn't one of
+    // `ANALYZER_IDS` — that list is fetch-fetch.ts's heavyweight analyzers
+    // only). It has three distinct states, all of which must render — the
+    // fix-round F2 finding: with `refreshRunners=cached` and no snapshot ever
+    // written yet, BOTH of the two notes below used to stay empty (one gated
+    // on `scannedAt` being present, the other on the "not requested" branch
+    // this call isn't in), so the original #1623 silence survived in exactly
+    // this mode. `cheapScanScannedAtMs` centralizes the one bit every branch
+    // needs: whether the stored snapshot has a usable timestamp (#1623
+    // fix-round F4 — a corrupt/missing `scannedAt` must not compute a NaN
+    // age, mirroring `./cache.ts`'s `Number.isFinite` guard).
+    const cheapScanScannedAtMs = rawProjectSnapshot?.scannedAt
+        ? Date.parse(rawProjectSnapshot.scannedAt)
+        : undefined;
+    const cheapScanStatusNote = !projectRunnersRequested
+        ? `\n\ncheap project scan (tree-sitter/fact-rules/ast-grep): not run this call (${NOT_REQUESTED_REASON}).`
+        : options.refreshRunners === "cached"
+            ? cheapScanScannedAtMs !== undefined && Number.isFinite(cheapScanScannedAtMs)
+                ? `\n\ncheap project scan (tree-sitter/fact-rules/ast-grep): served from cache, ${formatCacheAgeOld(Date.now() - cheapScanScannedAtMs)} — not re-run this call.`
+                : `\n\ncheap project scan (tree-sitter/fact-rules/ast-grep): not run (no cached scan; refresh with refreshRunners=cheap/all to populate).`
+            : "";
     const abortedNote = abortedIds.size > 0
         ? `\n\nstopped mid-scan (still running in the background, not reflected in this result): ${[
             ...abortedIds,
@@ -1208,12 +1533,25 @@ async function formatFullMode(cwd, severity, lspService, cacheManager, options =
     // possibly-stale session_start cache — say so honestly, with per-analyzer
     // elapsed time, since this can legitimately take a while (trivy's own
     // ~180s ceiling).
-    const timingEntries = Object.entries(extracted.timings ?? {});
+    // #1623: `extracted.cachedAgeMs` marks ids that are a cache-read BY DESIGN
+    // (today, only test-runner — see fresh-fetch.ts) rather than a fresh
+    // execution this call. Exclude those from "fetched fresh" below — folding
+    // a cache replay into the same list as a genuine run is exactly the
+    // misread the second #1623 dogfood pass reported (a cached result read as
+    // "just ran"). They get their own note with an actual age instead.
+    const cachedAgeMs = extracted.cachedAgeMs ?? {};
+    const timingEntries = Object.entries(extracted.timings ?? {}).filter(([id]) => !(id in cachedAgeMs));
     const freshNote = timingEntries.length > 0
         ? `\n\nfetched fresh this call: ${timingEntries
             .map(([id, ms]) => `${id} (${Math.round(ms)}ms)`)
             .join(", ")}.`
         : "";
+    const cachedAgeEntries = Object.entries(cachedAgeMs);
+    const cachedAgeNote = cachedAgeEntries.length > 0
+        ? `\n\nserved from cache this call (not re-run): ${cachedAgeEntries
+            .map(([id, ms]) => `${id} (${formatCacheAgeOld(ms)})`)
+            .join(", ")}.`
+        : "";
     // coldRunners always lands in details (even when empty) so a caller can
     // reliably check "were any extractors cold" without a presence check.
     // analyzerTimingsMs surfaces the #585 fresh-fetch elapsed time per
@@ -1224,8 +1562,19 @@ async function formatFullMode(cwd, severity, lspService, cacheManager, options =
         details: {
             ...result.details,
             coldRunners: extracted.cold,
+            // #1623: the specific reason each cold id was skipped (single source
+            // of truth: extractors.ts's `formatNotRunEntry` renders the same map
+            // into the text note above) — lets a caller check WHY without parsing
+            // text.
+            coldReasons: extracted.coldReasons ?? {},
             failedAnalyzers,
             analyzerTimingsMs: extracted.timings,
+            dispositionSuppressed: extracted.dispositionSuppressed ?? 0,
+            dispositionSuppressedByLane: extracted.dispositionSuppressedByLane ?? {},
+            // #1623: ms-old each cache-read-by-design lane's data was (today, only
+            // test-runner) — lets a caller distinguish "just ran" from "served
+            // from cache" without parsing the text note.
+            analyzersCachedAgeMs: cachedAgeMs,
             analyzersAborted: extracted.aborted ?? false,
             analyzersAbortedIds: extracted.abortedIds ?? [],
             // #747: true when the fresh fetch refused an at-or-above-$HOME root —
@@ -1241,6 +1590,8 @@ async function formatFullMode(cwd, severity, lspService, cacheManager, options =
             // any files unconfirmed" without re-deriving it from the text.
             lspFilesConfirmed: confirmedLspResults.length,
             lspFilesUnconfirmed: unconfirmedLspResults.length,
+            lspFilesPartiallyCovered: partiallyCoveredLspResults.length,
+            unconfirmedLspServerIds: uncoveredScannerIds,
             unconfirmedLspFiles: unconfirmedLspResults.map((result) => result.filePath),
             // #646: per-primary-server breakdown of the same tally, plus the
             // primary/auxiliary findings split — mirrors `lsp_diagnostics`'
@@ -1287,15 +1638,19 @@ async function formatFullMode(cwd, severity, lspService, cacheManager, options =
                     text: result.content[0].text +
                         note +
                         unconfirmedLspNote +
+                        auxiliaryCoverageNote +
                         lspPrimaryVsAuxiliaryNote +
                         coldNote +
                         testRunnerEditScopedNote +
                         failedNote +
+                        dispositionSuppressedNote +
                         walkUnsafeRootNote +
                         scanTruncatedNote +
                         generatedSkipNote +
                         abortedNote +
                         freshNote +
+                        cachedAgeNote +
+                        cheapScanStatusNote +
                         missingNote,
                 },
             ],
@@ -1306,12 +1661,16 @@ async function formatFullMode(cwd, severity, lspService, cacheManager, options =
         coldNote ||
         testRunnerEditScopedNote ||
         failedNote ||
+        dispositionSuppressedNote ||
         walkUnsafeRootNote ||
         scanTruncatedNote ||
         generatedSkipNote ||
         abortedNote ||
         freshNote ||
+        cachedAgeNote ||
+        cheapScanStatusNote ||
         unconfirmedLspNote ||
+        auxiliaryCoverageNote ||
         lspPrimaryVsAuxiliaryNote) {
         return {
             content: [
@@ -1319,15 +1678,19 @@ async function formatFullMode(cwd, severity, lspService, cacheManager, options =
                     type: "text",
                     text: result.content[0].text +
                         unconfirmedLspNote +
+                        auxiliaryCoverageNote +
                         lspPrimaryVsAuxiliaryNote +
                         coldNote +
                         testRunnerEditScopedNote +
                         failedNote +
+                        dispositionSuppressedNote +
                         walkUnsafeRootNote +
                         scanTruncatedNote +
                         generatedSkipNote +
                         abortedNote +
                         freshNote +
+                        cachedAgeNote +
+                        cheapScanStatusNote +
                         missingNote,
                 },
             ],
@@ -1336,13 +1699,17 @@ async function formatFullMode(cwd, severity, lspService, cacheManager, options =
     }
     return resultWithCold;
 }
-function formatAllMode(cwd, severity, summaries = getFileDiagnosticSummaries(), detailOverrides = { mode: "all" }, staleDropped = 0, pathsScope) {
+// @delivery-surface: lens-diagnostics:mode-all
+function formatAllMode(cwd, severity, summaries = getFileDiagnosticSummaries(), detailOverrides = { mode: "all" }, staleDropped = 0, pathsScope, dependencyDemoted = 0) {
     // Files changed/deleted since their diagnostics were recorded have already
     // been dropped by reconcileStaleWidgetFiles; note them so the agent knows
     // those aren't "clean", just un-rescanned (use mode=full to refresh).
-    const staleNote = staleDropped > 0
+    const staleNote = (staleDropped > 0
         ? ` (${staleDropped} changed file${staleDropped === 1 ? "" : "s"} omitted as stale — use mode=full to rescan)`
-        : "";
+        : "") +
+        (dependencyDemoted > 0
+            ? ` (${dependencyDemoted} blocking finding${dependencyDemoted === 1 ? "" : "s"} demoted to [stale] — an imported file changed since; re-run to confirm)`
+            : "");
     // mode=full already actively scanned exactly the requested paths, so a zero
     // result there IS a legitimate clean read — the cache-only note only applies
     // to delta/all, which is why this is gated on detailOverrides.mode !== "full".
@@ -1370,14 +1737,43 @@ function formatAllMode(cwd, severity, summaries = getFileDiagnosticSummaries(),
                 return s;
             return summarizeDiagnostics(s.filePath, policyKept, s.hasFinalSnapshot);
         });
-    const visibleSummaries = dispositioned.filter((s) => includeFile(s.filePath));
-    // Filter to files with actual issues
+    // #1641: the single chokepoint both mode=all (cache-only `summaries`) and
+    // mode=full (widget cache merged with a FRESH LSP sweep, above) converge
+    // through — a fresh sweep's server can still be citing its own stale
+    // in-memory document, so gating only the cached half would miss exactly
+    // the live incident this issue reports. Runs AFTER dispositions/rule
+    // policy (not before): a finding a mark or a `.pi-lens.json` rule policy
+    // already dropped is not being served, so it must never trigger a resync
+    // or a `diagnostic_past_eof` telemetry record on this call.
+    const eofGated = dispositioned.map((s) => {
+        const { diagnostics, demotedCount } = demotePastEofDiagnostics({
+            store: "lens_diagnostics",
+            cwd,
+            filePath: s.filePath,
+            diagnostics: s.diagnostics ?? [],
+            resync: resyncDocumentOnPastEof,
+        });
+        if (demotedCount === 0)
+            return s;
+        return summarizeDiagnostics(s.filePath, diagnostics, s.hasFinalSnapshot);
+    });
+    const visibleSummaries = eofGated.filter((s) => includeFile(s.filePath));
+    // Filter to files with actual issues. A file whose only findings were
+    // demoted by the #1641 past-EOF gate has blocking/errors/warnings at 0 —
+    // counting it as "issue-free" would tell the agent the file is CLEAN when
+    // it actually has an unconfirmed finding waiting on a rescan. Surface it
+    // under `severity: "all"` (the gate's own advisory tier) so it stays
+    // visible; a narrower `error`/`warning` filter still excludes it, matching
+    // how those filters already treat any other non-matching severity.
     const withIssues = visibleSummaries.filter((s) => {
         if (severity === "error")
             return s.blocking > 0 || s.errors > 0;
         if (severity === "warning")
             return s.warnings > 0;
-        return s.blocking > 0 || s.errors > 0 || s.warnings > 0;
+        return (s.blocking > 0 ||
+            s.errors > 0 ||
+            s.warnings > 0 ||
+            (s.diagnostics ?? []).some((d) => d.stale));
     });
     const pathsNote = isFullMode ? "" : pathsScopeCacheOnlyNote(pathsScope);
     if (withIssues.length === 0) {
@@ -1412,6 +1808,10 @@ function formatAllMode(cwd, severity, summaries = getFileDiagnosticSummaries(),
             parts.push(`${s.warnings}W`);
         if (!s.hasFinalSnapshot)
             parts.push(`(pending)`);
+        const staleCount = (s.diagnostics ?? []).filter((d) => d.stale).length;
+        if (staleCount > 0) {
+            parts.push(`${staleCount} stale — re-run to confirm`);
+        }
         lines.push(`${rel}  ${parts.join("  ")}`);
         // List the actual diagnostics (not just counts) so the agent can act on
         // them without re-running anything — same "L<line>: <message>" shape as the
@@ -1423,16 +1823,27 @@ function formatAllMode(cwd, severity, summaries = getFileDiagnosticSummaries(),
             .sort(bySeverityThenLine);
         const shown = matching.slice(0, MAX_DIAGNOSTICS_PER_FILE);
         for (const d of shown) {
-            const marker = isErrorLike(d)
-                ? d.semantic === "blocking"
-                    ? "🔴 "
-                    : ""
-                : "";
+            const marker = d.stale
+                ? ""
+                : isErrorLike(d)
+                    ? d.semantic === "blocking"
+                        ? "🔴 "
+                        : ""
+                    : "";
             const label = d.rule ?? d.tool;
             const tag = label ? ` [${label}]` : "";
             const flaggedTag = d.flagged ? " 📌 flagged-to-fix" : "";
             const msg = d.message.replace(/\s+/g, " ").trim();
-            lines.push(`  ${marker}L${d.line ?? "?"}: ${msg}${tag}${flaggedTag}`);
+            // #1641: a demoted past-EOF entry no longer has a trustworthy line —
+            // show the marker instead of a coordinate that doesn't exist in the
+            // current file, matching the #1622/#1627 stale-line render convention.
+            // Past-EOF demotions replace the untrustworthy coordinate; other
+            // demotions (#1631 drift) keep their in-bounds line and append the
+            // shared stale marker instead.
+            const pastEof = d.stale && (d.staleReason ?? "past-eof") === "past-eof";
+            const loc = pastEof ? PAST_EOF_STALE_MARKER : `L${d.line ?? "?"}`;
+            const staleTag = d.stale && !pastEof ? ` ${STALE_LINE_MARKER}` : "";
+            lines.push(`  ${marker}${loc}: ${msg}${tag}${flaggedTag}${staleTag}`);
         }
         if (matching.length > shown.length) {
             lines.push(`  … ${matching.length - shown.length} more in this file (showing ${shown.length} of ${matching.length})`);
@@ -1441,12 +1852,20 @@ function formatAllMode(cwd, severity, summaries = getFileDiagnosticSummaries(),
         totalErrors += s.errors;
         totalWarnings += s.warnings;
     }
+    // #1799: `totalBlocking` and `totalErrors` count the same findings UNLESS a
+    // dependency-drift demotion (#1631) has revoked blocking authority from an
+    // error while leaving it in the error tally (widget-state.ts `isBlocking`
+    // vs `countDiagnostics`) — that's a real, live disagreement between the two
+    // totals, not double-counting. So `errors` renders only when
+    // `blocking === 0`, same guard as the per-file row above: it shows the
+    // drift-demoted findings the blocking line can't, without ever printing the
+    // same findings twice under two labels when blocking > 0.
     const summary = [
         `\nSummary (${visibleSummaries.length} files diagnosed this session):`,
         totalBlocking > 0
             ? `  🔴 ${totalBlocking} blocking error${totalBlocking === 1 ? "" : "s"}`
             : null,
-        totalErrors > 0
+        totalBlocking === 0 && totalErrors > 0
             ? `  ${totalErrors} error${totalErrors === 1 ? "" : "s"}`
             : null,
         totalWarnings > 0