@aztec/p2p 0.0.1-commit.35158ae7e → 0.0.1-commit.3750d92a7

package/src/services/reqresp/protocols/tx.ts

@@ -51,9 +51,7 @@ export function reqRespTxHandler(mempools: MemPools): ReqRespSubProtocolHandler
  *  Per: https://github.com/AztecProtocol/aztec-packages/issues/15149#issuecomment-2999054485
  *  we define Q as max number of transactions per batch, the comment explains why we use 8.
  */
-//TODO: (mralj) chunk size should by default be 8, this is just temporary until the protocol is implemented correctly
-//more info:  https://github.com/AztecProtocol/aztec-packages/pull/15516#pullrequestreview-2995474321
-export function chunkTxHashesRequest(hashes: TxHash[], chunkSize = 1): Array<TxHashArray> {
+export function chunkTxHashesRequest(hashes: TxHash[], chunkSize = 8): Array<TxHashArray> {
   return chunk(hashes, chunkSize).map(chunk => new TxHashArray(...chunk));
 }
 

package/src/services/reqresp/rate-limiter/rate_limiter.ts

@@ -97,9 +97,10 @@ export function prettyPrintRateLimitStatus(status: RateLimitStatus) {
  * 2. Individual rate limits for each peer.
  *
  * How it works:
- * - When a request comes in, it first checks against the global rate limit.
- * - If the global limit allows, it then checks against the specific peer's rate limit.
- * - The request is only allowed if both the global and peer-specific limits allow it.
+ * - When a request comes in, it first checks against the peer's individual rate limit.
+ * - If the peer limit allows, it then checks against the global rate limit.
+ * - The request is only allowed if both the peer-specific and global limits allow it.
+ * - Checking peer limit first ensures a rate-limited peer cannot exhaust the global quota.
  * - It automatically creates and manages rate limiters for new peers as they make requests.
  * - It periodically cleans up rate limiters for inactive peers to conserve memory.
  *
@@ -119,10 +120,6 @@ export class SubProtocolRateLimiter {
   }
 
   allow(peerId: PeerId): RateLimitStatus {
-    if (!this.globalLimiter.allow()) {
-      return RateLimitStatus.DeniedGlobal;
-    }
-
     const peerIdStr = peerId.toString();
     let peerLimiter: PeerRateLimiter | undefined = this.peerLimiters.get(peerIdStr);
     if (!peerLimiter) {
@@ -135,10 +132,17 @@ export class SubProtocolRateLimiter {
     } else {
       peerLimiter.lastAccess = Date.now();
     }
-    const peerLimitAllowed = peerLimiter.limiter.allow();
-    if (!peerLimitAllowed) {
+
+    // Check peer limit first: a rate-limited peer must not consume global quota,
+    // otherwise one spamming peer can starve all others by exhausting the global bucket.
+    if (!peerLimiter.limiter.allow()) {
       return RateLimitStatus.DeniedPeer;
     }
+
+    if (!this.globalLimiter.allow()) {
+      return RateLimitStatus.DeniedGlobal;
+    }
+
     return RateLimitStatus.Allowed;
   }
 

package/src/services/reqresp/rate-limiter/rate_limits.ts

@@ -42,16 +42,6 @@ export const DEFAULT_RATE_LIMITS: ReqRespSubProtocolRateLimits = {
       quotaCount: 200,
     },
   },
-  [ReqRespSubProtocol.BLOCK]: {
-    peerLimit: {
-      quotaTimeMs: 1000,
-      quotaCount: 2,
-    },
-    globalLimit: {
-      quotaTimeMs: 1000,
-      quotaCount: 5,
-    },
-  },
   [ReqRespSubProtocol.GOODBYE]: {
     peerLimit: {
       quotaTimeMs: 1000,

package/src/services/reqresp/reqresp.ts

@@ -34,7 +34,9 @@ import {
   type ReqRespSubProtocolHandlers,
   type ReqRespSubProtocolRateLimits,
   type ReqRespSubProtocolValidators,
+  type ShouldRejectPeer,
   type SubProtocolMap,
+  UNAUTHENTICATED_ALLOWED_PROTOCOLS,
   responseFromBuffer,
   subProtocolSizeCalculators,
 } from './interface.js';
@@ -72,6 +74,8 @@ export class ReqResp implements ReqRespInterface {
 
   private snappyTransform: SnappyTransform;
 
+  private shouldRejectPeer: ShouldRejectPeer | undefined;
+
   private metrics: ReqRespMetrics;
 
   constructor(
@@ -108,6 +112,10 @@ export class ReqResp implements ReqRespInterface {
     }
   }
 
+  public setShouldRejectPeer(checker: ShouldRejectPeer): void {
+    this.shouldRejectPeer = checker;
+  }
+
   get tracer() {
     return this.metrics.tracer;
   }
@@ -462,7 +470,7 @@ export class ReqResp implements ReqRespInterface {
       );
       return resp;
     } catch (e: any) {
-      this.logger.warn(`SUBPROTOCOL: ${subProtocol}\n`, e);
+      this.logger.debug(`SUBPROTOCOL: ${subProtocol}\n`, e);
       // On error we immediately abort the stream, this is preferred way,
       // because it signals to the sender that error happened, whereas
       // closing the stream only closes our side and is much slower
@@ -596,6 +604,15 @@ export class ReqResp implements ReqRespInterface {
         throw new ReqRespStatusError(ReqRespStatus.RATE_LIMIT_EXCEEDED);
       }
 
+      // When p2pAllowOnlyValidators is enabled, reject unauthenticated peers on data protocols
+      if (
+        !UNAUTHENTICATED_ALLOWED_PROTOCOLS.has(protocol) &&
+        (this.shouldRejectPeer?.(connection.remotePeer.toString()) ?? false)
+      ) {
+        this.logger.debug(`Rejecting unauthenticated peer ${connection.remotePeer} on gated protocol ${protocol}`);
+        throw new ReqRespStatusError(ReqRespStatus.FAILURE);
+      }
+
       await this.processStream(protocol, incomingStream);
     } catch (err: any) {
       this.metrics.recordResponseError(protocol);

package/src/services/service.ts

@@ -117,7 +117,12 @@ export interface P2PService {
   // Leaky abstraction: fix https://github.com/AztecProtocol/aztec-packages/issues/7963
   registerBlockReceivedCallback(callback: P2PBlockReceivedCallback): void;
 
-  registerCheckpointReceivedCallback(callback: P2PCheckpointReceivedCallback): void;
+  registerValidatorCheckpointReceivedCallback(callback: P2PCheckpointReceivedCallback): void;
+
+  registerAllNodesCheckpointReceivedCallback(callback: P2PCheckpointReceivedCallback): void;
+
+  /** Fires the all-nodes checkpoint callback for our own proposal (gossipsub doesn't deliver own messages). */
+  notifyOwnCheckpointProposal(checkpoint: CheckpointProposalCore): Promise<void>;
 
   /**
    * Registers a callback invoked when a duplicate proposal is detected (equivocation).
@@ -196,13 +201,6 @@ export interface PeerDiscoveryService extends EventEmitter {
   on(event: 'peer:discovered', listener: (enr: ENR) => void): this;
   emit(event: 'peer:discovered', enr: ENR): boolean;
 
-  /**
-   * Event emitted when our public IP is discovered or changes via discv5 peer interactions.
-   * Only emitted when enrUpdate is enabled (i.e. queryForIp=true and no static p2pIp).
-   */
-  on(event: 'ip:changed', listener: (ip: string) => void): this;
-  emit(event: 'ip:changed', ip: string): boolean;
-
   getStatus(): PeerDiscoveryState;
 
   getEnr(): ENR | undefined;

package/src/services/tx_collection/file_store_tx_source.ts

@@ -1,7 +1,8 @@
+import { partitionAsync } from '@aztec/foundation/collection';
 import { type Logger, createLogger } from '@aztec/foundation/log';
 import { Timer } from '@aztec/foundation/timer';
 import { type ReadOnlyFileStore, createReadOnlyFileStore } from '@aztec/stdlib/file-store';
-import { Tx, type TxHash } from '@aztec/stdlib/tx';
+import { Tx, type TxHash, type TxValidator } from '@aztec/stdlib/tx';
 import {
   type Histogram,
   Metrics,
@@ -23,6 +24,7 @@ export class FileStoreTxSource implements TxSource {
     private readonly fileStore: ReadOnlyFileStore,
     private readonly baseUrl: string,
     private readonly basePath: string,
+    private readonly txValidator: TxValidator,
     private readonly log: Logger,
     telemetry: TelemetryClient,
   ) {
@@ -44,6 +46,7 @@ export class FileStoreTxSource implements TxSource {
   public static async create(
     url: string,
     basePath: string,
+    txValidator: TxValidator,
     log: Logger = createLogger('p2p:file_store_tx_source'),
     telemetry: TelemetryClient = getTelemetryClient(),
   ): Promise<FileStoreTxSource | undefined> {
@@ -53,7 +56,7 @@ export class FileStoreTxSource implements TxSource {
         log.warn(`Failed to create file store for URL: ${url}`);
         return undefined;
       }
-      return new FileStoreTxSource(fileStore, url, basePath, log, telemetry);
+      return new FileStoreTxSource(fileStore, url, basePath, txValidator, log, telemetry);
     } catch (err) {
       log.warn(`Error creating file store for URL: ${url}`, { error: err });
       return undefined;
@@ -65,35 +68,41 @@ export class FileStoreTxSource implements TxSource {
   }
 
   public async getTxsByHash(txHashes: TxHash[]): Promise<TxSourceCollectionResult> {
-    const invalidTxHashes: string[] = [];
+    const results = await Promise.all(
+      txHashes.map(async txHash => {
+        const path = `${this.basePath}/txs/${txHash.toString()}.bin`;
+        const timer = new Timer();
+        try {
+          const buffer = await this.fileStore.read(path);
+          const tx = Tx.fromBuffer(buffer);
+          return { tx, downloadDuration: timer.ms(), downloadSize: buffer.length };
+        } catch {
+          this.downloadsFailed.add(1);
+          return undefined;
+        }
+      }),
+    );
+
+    const txs = results.filter(tx => tx !== undefined);
+    const [validTxs, invalidTxs] = await partitionAsync(
+      txs,
+      async ({ tx, downloadDuration, downloadSize }): Promise<boolean> => {
+        const valid = await this.txValidator.validateTx(tx);
+        if (valid.result === 'valid') {
+          this.downloadsSuccess.add(1);
+          this.downloadDuration.record(Math.ceil(downloadDuration));
+          this.downloadSize.record(downloadSize);
+          return true;
+        } else {
+          this.downloadsFailed.add(1);
+          return false;
+        }
+      },
+    );
+
     return {
-      validTxs: (
-        await Promise.all(
-          txHashes.map(async txHash => {
-            const path = `${this.basePath}/txs/${txHash.toString()}.bin`;
-            const timer = new Timer();
-            try {
-              const buffer = await this.fileStore.read(path);
-              const tx = Tx.fromBuffer(buffer);
-              if ((await tx.validateTxHash()) && txHash.equals(tx.txHash)) {
-                this.downloadsSuccess.add(1);
-                this.downloadDuration.record(Math.ceil(timer.ms()));
-                this.downloadSize.record(buffer.length);
-                return tx;
-              } else {
-                invalidTxHashes.push(tx.txHash.toString());
-                this.downloadsFailed.add(1);
-                return undefined;
-              }
-            } catch {
-              // Tx not found or error reading - return undefined
-              this.downloadsFailed.add(1);
-              return undefined;
-            }
-          }),
-        )
-      ).filter(tx => tx !== undefined),
-      invalidTxHashes: invalidTxHashes,
+      validTxs: validTxs.map(({ tx }) => tx),
+      invalidTxHashes: invalidTxs.map(({ tx }) => tx.getTxHash().toString()),
     };
   }
 }
@@ -109,9 +118,12 @@ export class FileStoreTxSource implements TxSource {
 export async function createFileStoreTxSources(
   urls: string[],
   basePath: string,
+  txValidator: TxValidator,
   log: Logger = createLogger('p2p:file_store_tx_source'),
   telemetry: TelemetryClient = getTelemetryClient(),
 ): Promise<FileStoreTxSource[]> {
-  const sources = await Promise.all(urls.map(url => FileStoreTxSource.create(url, basePath, log, telemetry)));
+  const sources = await Promise.all(
+    urls.map(url => FileStoreTxSource.create(url, basePath, txValidator, log, telemetry)),
+  );
   return sources.filter((s): s is FileStoreTxSource => s !== undefined);
 }

package/src/services/tx_collection/tx_source.ts

@@ -2,7 +2,7 @@ import { getVKTreeRoot } from '@aztec/noir-protocol-circuits-types/vk-tree';
 import { protocolContractsHash } from '@aztec/protocol-contracts';
 import type { ChainConfig } from '@aztec/stdlib/config';
 import { type AztecNode, createAztecNodeClient } from '@aztec/stdlib/interfaces/client';
-import type { Tx, TxHash } from '@aztec/stdlib/tx';
+import type { Tx, TxHash, TxValidator } from '@aztec/stdlib/tx';
 import { type ComponentsVersions, getComponentsVersionsFromConfig } from '@aztec/stdlib/versioning';
 import { makeTracedFetch } from '@aztec/telemetry-client';
 
@@ -16,12 +16,13 @@ export interface TxSource {
 export class NodeRpcTxSource implements TxSource {
   constructor(
     private readonly client: Pick<AztecNode, 'getTxsByHash'>,
+    private readonly txValidator: TxValidator,
     private readonly info: string,
   ) {}
 
-  public static fromUrl(nodeUrl: string, versions: ComponentsVersions): NodeRpcTxSource {
+  public static fromUrl(nodeUrl: string, txValidator: TxValidator, versions: ComponentsVersions): NodeRpcTxSource {
     const client = createAztecNodeClient(nodeUrl, versions, makeTracedFetch([1, 2, 3], false));
-    return new NodeRpcTxSource(client, nodeUrl);
+    return new NodeRpcTxSource(client, txValidator, nodeUrl);
   }
 
   public getInfo() {
@@ -38,8 +39,8 @@ export class NodeRpcTxSource implements TxSource {
     const invalidTxHashes: string[] = [];
     await Promise.all(
       txs.map(async tx => {
-        const isValid = await tx.validateTxHash();
-        if (isValid) {
+        const validation = await this.txValidator.validateTx(tx);
+        if (validation.result === 'valid') {
           validTxs.push(tx);
         } else {
           invalidTxHashes.push(tx.getTxHash().toString());
@@ -50,7 +51,7 @@ export class NodeRpcTxSource implements TxSource {
   }
 }
 
-export function createNodeRpcTxSources(urls: string[], chainConfig: ChainConfig) {
+export function createNodeRpcTxSources(urls: string[], txValidator: TxValidator, chainConfig: ChainConfig) {
   const versions = getComponentsVersionsFromConfig(chainConfig, protocolContractsHash, getVKTreeRoot());
-  return urls.map(url => NodeRpcTxSource.fromUrl(url, versions));
+  return urls.map(url => NodeRpcTxSource.fromUrl(url, txValidator, versions));
 }

package/src/test-helpers/mock-pubsub.ts

@@ -1,5 +1,6 @@
 import type { EpochCacheInterface } from '@aztec/epoch-cache';
 import { type Logger, createLogger } from '@aztec/foundation/log';
+import { sleep } from '@aztec/foundation/sleep';
 import type { AztecAsyncKVStore } from '@aztec/kv-store';
 import type { L2BlockSource } from '@aztec/stdlib/block';
 import type { ContractDataSource } from '@aztec/stdlib/contract';
@@ -15,7 +16,6 @@ import {
   type TopicValidatorResult,
   TypedEventEmitter,
 } from '@libp2p/interface';
-import type { AddressManager, ConnectionManager } from '@libp2p/interface-internal';
 
 import type { P2PConfig } from '../config.js';
 import type { MemPools } from '../mem_pools/interface.js';
@@ -101,6 +101,7 @@ class MockReqResp implements ReqRespInterface {
   }
 
   updateConfig(_config: Partial<P2PReqRespConfig>): void {}
+  setShouldRejectPeer(): void {}
 
   start(
     subProtocolHandlers: Partial<ReqRespSubProtocolHandlers>,
@@ -139,6 +140,11 @@ class MockReqResp implements ReqRespInterface {
     const responses: InstanceType<SubProtocolMap[SubProtocol]['response']>[] = [];
     const peers = this.network.getReqRespPeers().filter(p => !p.peerId.equals(this.peerId));
     const targetPeers = pinnedPeer ? peers.filter(p => p.peerId.equals(pinnedPeer)) : peers;
+    const delayMs = this.network.getPropagationDelayMs();
+
+    if (delayMs > 0) {
+      await sleep(delayMs);
+    }
 
     for (const request of requests) {
       const requestBuffer = request.toBuffer();
@@ -175,7 +181,12 @@ class MockReqResp implements ReqRespInterface {
       return { status: ReqRespStatus.SUCCESS, data: Buffer.from([]) };
     }
     try {
+      const delayMs = this.network.getPropagationDelayMs();
+      if (delayMs > 0) {
+        await sleep(delayMs);
+      }
       const data = await handler(this.peerId, payload);
+
       return { status: ReqRespStatus.SUCCESS, data };
     } catch {
       return { status: ReqRespStatus.FAILURE };
@@ -213,14 +224,6 @@ export class MockPubSub implements PubSubLibp2p {
   get services() {
     return {
       pubsub: this.gossipSub,
-      components: {
-        addressManager: {
-          removeObservedAddr: () => {},
-          addObservedAddr: () => {},
-          confirmObservedAddr: () => {},
-        } as unknown as AddressManager,
-        connectionManager: {} as unknown as ConnectionManager,
-      },
     };
   }
 
@@ -251,10 +254,10 @@ class MockGossipSubService extends TypedEventEmitter<GossipsubEvents> implements
     score: (_peerId: PeerIdStr) => 0,
   };
 
-  publish(topic: TopicStr, data: Uint8Array, _opts?: PublishOpts): Promise<PublishResult> {
+  async publish(topic: TopicStr, data: Uint8Array, _opts?: PublishOpts): Promise<PublishResult> {
     this.logger.debug(`Publishing message on topic ${topic}`, { topic, sender: this.peerId.toString() });
-    this.network.publishToPeers(topic, data, this.peerId);
-    return Promise.resolve({ recipients: this.network.getPeers().filter(peer => !this.peerId.equals(peer)) });
+    await this.network.publishToPeers(topic, data, this.peerId);
+    return { recipients: this.network.getPeers().filter(peer => !this.peerId.equals(peer)) };
   }
 
   receive(msg: GossipsubMessage) {
@@ -290,7 +293,8 @@ class MockGossipSubService extends TypedEventEmitter<GossipsubEvents> implements
 
 /**
  * Mock gossip sub network used for testing.
- * All instances of MockGossipSubService connected to the same network will instantly receive the same messages.
+ * All instances of MockGossipSubService connected to the same network receive the same messages,
+ * optionally delayed by a configurable propagation time.
  */
 export class MockGossipSubNetwork {
   private peers: MockGossipSubService[] = [];
@@ -299,6 +303,15 @@ export class MockGossipSubNetwork {
 
   private logger = createLogger('p2p:test:mock-gossipsub-network');
 
+  constructor(
+    /** Artificial propagation delay in milliseconds applied to each message delivery. */
+    private propagationDelayMs: number = 0,
+  ) {}
+
+  public getPropagationDelayMs(): number {
+    return this.propagationDelayMs;
+  }
+
   public getPeers(): PeerId[] {
     return this.peers.map(peer => peer.peerId);
   }
@@ -315,7 +328,7 @@ export class MockGossipSubNetwork {
     return this.reqRespPeers;
   }
 
-  public publishToPeers(topic: TopicStr, data: Uint8Array, sender: PeerId): void {
+  public async publishToPeers(topic: TopicStr, data: Uint8Array, sender: PeerId): Promise<void> {
     const msgId = (this.nextMsgId++).toString();
     this.logger.debug(`Network is distributing message on topic ${topic}`, {
       topic,
@@ -324,6 +337,10 @@ export class MockGossipSubNetwork {
       msgId,
     });
 
+    if (this.propagationDelayMs > 0) {
+      await sleep(this.propagationDelayMs);
+    }
+
     const gossipSubMsg: GossipsubMessage = { msgId, msg: { type: 'unsigned', topic, data }, propagationSource: sender };
     for (const peer of this.peers) {
       if (peer.subscribedTopics.has(topic)) {

package/src/test-helpers/reqresp-nodes.ts

@@ -153,6 +153,8 @@ export async function createTestLibP2PService(
     epochCache,
   );
 
+  reqresp.setShouldRejectPeer(peerId => peerManager.shouldDisableP2PGossip(peerId));
+
   p2pNode.services.pubsub.score.params.appSpecificWeight = APP_SPECIFIC_WEIGHT;
   p2pNode.services.pubsub.score.params.appSpecificScore = (peerId: string) =>
     peerManager.shouldDisableP2PGossip(peerId) ? -Infinity : peerManager.getPeerScore(peerId);
@@ -187,7 +189,6 @@ export const MOCK_SUB_PROTOCOL_HANDLERS: ReqRespSubProtocolHandlers = {
   [ReqRespSubProtocol.STATUS]: (_msg: any) => Promise.resolve(Buffer.from('status')),
   [ReqRespSubProtocol.TX]: (_msg: any) => Promise.resolve(Buffer.from('tx')),
   [ReqRespSubProtocol.GOODBYE]: (_msg: any) => Promise.resolve(Buffer.from('goodbye')),
-  [ReqRespSubProtocol.BLOCK]: (_msg: any) => Promise.resolve(Buffer.from('block')),
   [ReqRespSubProtocol.AUTH]: (_msg: any) => Promise.resolve(Buffer.from('auth')),
   [ReqRespSubProtocol.BLOCK_TXS]: (_msg: any) => Promise.resolve(Buffer.from('block_txs')),
 };
@@ -199,7 +200,6 @@ export const MOCK_SUB_PROTOCOL_VALIDATORS: ReqRespSubProtocolValidators = {
   [ReqRespSubProtocol.STATUS]: noopValidator,
   [ReqRespSubProtocol.TX]: noopValidator,
   [ReqRespSubProtocol.GOODBYE]: noopValidator,
-  [ReqRespSubProtocol.BLOCK]: noopValidator,
   [ReqRespSubProtocol.AUTH]: noopValidator,
   [ReqRespSubProtocol.BLOCK_TXS]: noopValidator,
 };

package/src/test-helpers/testbench-utils.ts

@@ -287,6 +287,7 @@ export function createMockEpochCache(): EpochCacheInterface {
       nowMs: 0n,
     }),
     isProposerPipeliningEnabled: () => false,
+    pipeliningOffset: () => 0,
     computeProposerIndex: () => 0n,
     getCurrentAndNextSlot: () => ({ currentSlot: SlotNumber.ZERO, nextSlot: SlotNumber.ZERO }),
     getTargetAndNextSlot: () => ({ targetSlot: SlotNumber.ZERO, nextSlot: SlotNumber.ZERO }),

package/src/testbench/p2p_client_testbench_worker.ts

@@ -204,6 +204,25 @@ function installUnlimitedRateLimits(client: P2PClient): void {
   rateLimiter.allow = () => RateLimitStatus.Allowed;
 }
 
+/** Resets peer scores to prevent cross-case contamination in benchmarks. */
+function resetPeerScores(client: P2PClient): void {
+  const peerManager = (client as any).p2pService.peerManager;
+  const peerScoring = peerManager?.peerScoring;
+  if (peerScoring?.resetAllScores) {
+    peerScoring.resetAllScores();
+  }
+}
+
+/** Returns the number of connected peers for connectivity checks. */
+function getConnectedPeerCount(client: P2PClient): number {
+  const p2pService = (client as any).p2pService;
+  const connectionSampler = p2pService?.reqresp?.getConnectionSampler?.();
+  if (connectionSampler?.getPeerListSortedByConnectionCountAsc) {
+    return connectionSampler.getPeerListSortedByConnectionCountAsc().length;
+  }
+  return 0;
+}
+
 async function runAggregatorBenchmark(
   client: P2PClient,
   blockProposal: BlockProposal,
@@ -321,6 +340,37 @@ let workerConfig: P2PConfig | null = null;
 let workerLogger: Logger | null = null;
 let kvStore: Awaited<ReturnType<typeof openTmpStore>> | null = null;
 
+async function stopWorker() {
+  try {
+    if (workerClient) {
+      await workerClient.stop();
+      workerClient = null;
+    }
+  } catch (e) {
+    workerLogger?.error('Error stopping worker client', e);
+  }
+  try {
+    if (kvStore?.close) {
+      await kvStore.close();
+      kvStore = null;
+    }
+  } catch (e) {
+    workerLogger?.error('Error closing kv store', e);
+  }
+}
+
+function gracefulExit(code: number = 0) {
+  try {
+    if (process.connected) {
+      process.disconnect();
+    }
+  } catch {
+    // IPC channel already closed
+  }
+  // Safety fallback if lingering handles prevent the event loop from draining
+  setTimeout(() => process.exit(code), 5000).unref();
+}
+
 // eslint-disable-next-line @typescript-eslint/no-misused-promises
 process.on('message', async msg => {
   const {
@@ -410,13 +460,8 @@ process.on('message', async msg => {
     const cmd = msg as any;
     switch (cmd.type) {
       case 'STOP':
-        if (workerClient) {
-          await workerClient.stop();
-        }
-        if (kvStore?.close) {
-          await kvStore.close();
-        }
-        process.exit(0);
+        await stopWorker();
+        gracefulExit(0);
         break;
 
       case 'SEND_TX':
@@ -426,6 +471,13 @@ process.on('message', async msg => {
         }
         break;
 
+      case 'GET_PEER_COUNT':
+        process.send!({
+          type: 'PEER_COUNT',
+          count: workerClient ? getConnectedPeerCount(workerClient) : 0,
+        });
+        break;
+
       case 'BENCH_REQRESP': {
         const benchCmd = cmd as BenchReqRespCommand;
         if (!workerClient || !workerTxPool || !workerAttestationPool || !workerConfig || !workerLogger) {
@@ -442,6 +494,7 @@ process.on('message', async msg => {
         // Reset state before each benchmark run to avoid cross-run contamination
         workerTxPool.resetState();
         workerAttestationPool.resetState();
+        resetPeerScores(workerClient);
 
         installUnlimitedRateLimits(workerClient);
 
@@ -493,7 +546,12 @@ process.on('message', async msg => {
       }
     }
   } catch (err: any) {
-    process.send!({ type: 'ERROR', error: err.message });
-    process.exit(1);
+    try {
+      process.send!({ type: 'ERROR', error: err.message });
+    } catch {
+      // IPC channel may be closed
+    }
+    await stopWorker();
+    gracefulExit(1);
   }
 });

package/src/testbench/worker_client_manager.ts

@@ -72,7 +72,6 @@ class WorkerClientManager {
   destroy() {
     this.cleanup().catch((error: Error) => {
       this.logger.error('Failed to cleanup worker client manager', error);
-      process.exit(1);
     });
   }
 
@@ -425,6 +424,61 @@ class WorkerClientManager {
     this.logger.info('All worker processes cleaned up');
   }
 
+  /**
+   * Checks that the aggregator (client 0) has sufficient peer connections before running a benchmark.
+   * This prevents benchmark cases from starting with degraded connectivity after a previous case
+   * caused connection failures.
+   */
+  async waitForConnectivity(minPeers: number, timeoutMs: number = 15_000): Promise<number> {
+    const waitInterval = 1000;
+    let waited = 0;
+
+    while (waited < timeoutMs) {
+      const count = await this.getPeerCount(0, 5000);
+      if (count >= minPeers) {
+        this.logger.info(`Connectivity check passed: ${count}/${minPeers} peers connected`);
+        return count;
+      }
+      this.logger.debug(`Waiting for connectivity: ${count}/${minPeers} (waited ${waited}ms)`);
+      await sleep(waitInterval);
+      waited += waitInterval;
+    }
+
+    const finalCount = await this.getPeerCount(0, 5000);
+    this.logger.warn(`Connectivity check: only ${finalCount}/${minPeers} peers after ${timeoutMs}ms`);
+    return finalCount;
+  }
+
+  private getPeerCount(clientIndex: number, timeoutMs: number): Promise<number> {
+    return new Promise<number>(resolve => {
+      let resolved = false;
+
+      const handler = (msg: any) => {
+        if (resolved) {
+          return;
+        }
+        if (msg.type === 'PEER_COUNT') {
+          resolved = true;
+          clearTimeout(timeout);
+          this.processes[clientIndex].off('message', handler);
+          resolve(msg.count as number);
+        }
+      };
+
+      const timeout = setTimeout(() => {
+        if (resolved) {
+          return;
+        }
+        resolved = true;
+        this.processes[clientIndex].off('message', handler);
+        resolve(0);
+      }, timeoutMs);
+
+      this.processes[clientIndex].on('message', handler);
+      this.processes[clientIndex].send({ type: 'GET_PEER_COUNT' });
+    });
+  }
+
   /**
    * Run a req/resp benchmark across all worker clients.
    *