@aztec/prover-client 0.0.1-commit.e558bd1c → 0.0.1-commit.e5a3663dd

package/src/mocks/test_context.ts

@@ -24,6 +24,7 @@ import {
   StateReference,
   TreeSnapshots,
 } from '@aztec/stdlib/tx';
+import type { GenesisData } from '@aztec/stdlib/world-state';
 import type { MerkleTreeAdminDatabase } from '@aztec/world-state';
 import { NativeWorldStateService } from '@aztec/world-state/native';
 
@@ -84,14 +85,13 @@ export class TestContext {
     const feePayer = AztecAddress.fromNumber(42222);
     const initialFeePayerBalance = new Fr(10n ** 20n);
     const feePayerSlot = await computeFeePayerBalanceLeafSlot(feePayer);
-    const prefilledPublicData = [new PublicDataTreeLeaf(feePayerSlot, initialFeePayerBalance)];
+    const genesis: GenesisData = {
+      prefilledPublicData: [new PublicDataTreeLeaf(feePayerSlot, initialFeePayerBalance)],
+      genesisTimestamp: 0n,
+    };
 
     // Separated dbs for public processor and prover - see public_processor for context
-    const ws = await NativeWorldStateService.tmp(
-      /*rollupAddress=*/ undefined,
-      /*cleanupTmpDir=*/ true,
-      prefilledPublicData,
-    );
+    const ws = await NativeWorldStateService.tmp(/*rollupAddress=*/ undefined, /*cleanupTmpDir=*/ true, genesis);
 
     let localProver: ServerCircuitProver;
     const config = await getEnvironmentConfig(logger);
@@ -104,8 +104,10 @@ export class TestContext {
         bbBinaryPath: config.expectedBBPath,
         bbWorkingDirectory: config.bbWorkingDirectory,
         bbSkipCleanup: config.bbSkipCleanup,
-        numConcurrentIVCVerifiers: 2,
+        numConcurrentIVCVerifiers: 8,
         bbIVCConcurrency: 1,
+        bbChonkVerifyMaxBatch: 16,
+        bbChonkVerifyConcurrency: 6,
       };
       localProver = await createProver(bbConfig);
     }
@@ -116,7 +118,7 @@ export class TestContext {
 
     const broker = new TestBroker(proverCount, localProver);
     const facade = new BrokerCircuitProverFacade(broker);
-    const orchestrator = new TestProvingOrchestrator(ws, facade, EthAddress.ZERO);
+    const orchestrator = new TestProvingOrchestrator(ws, facade, EthAddress.ZERO, false, 10);
 
     await broker.start();
     facade.start();
@@ -250,7 +252,7 @@ export class TestContext {
     const previousCheckpointOutHashes = this.checkpointOutHashes;
     const builder = await LightweightCheckpointBuilder.startNewCheckpoint(
       checkpointNumber,
-      constants,
+      { ...constants, timestamp },
       l1ToL2Messages,
       previousCheckpointOutHashes,
       cleanFork,
@@ -262,7 +264,7 @@ export class TestContext {
       const txs = blockTxs[i];
       const state = blockEndStates[i];
 
-      const block = await builder.addBlock(blockGlobalVariables[i], txs, {
+      const { block } = await builder.addBlock(blockGlobalVariables[i], txs, {
         expectedEndState: state,
         insertTxsEffects: true,
       });

package/src/orchestrator/block-building-helpers.ts

@@ -253,8 +253,8 @@ export function getPublicChonkVerifierPrivateInputsFromTx(tx: Tx | ProcessedTx,
 // Build "hints" as the private inputs for the checkpoint root rollup circuit.
 // The `blobCommitments` will be accumulated and checked in the root rollup against the `finalBlobChallenges`.
 // The `blobsHash` will be validated on L1 against the submitted blob data.
-export const buildBlobHints = (blobFields: Fr[]) => {
-  const blobs = getBlobsPerL1Block(blobFields);
+export const buildBlobHints = async (blobFields: Fr[]) => {
+  const blobs = await getBlobsPerL1Block(blobFields);
   const blobCommitments = getBlobCommitmentsFromBlobs(blobs);
   const blobsHash = computeBlobsHashFromBlobs(blobs);
   return { blobCommitments, blobs, blobsHash };

package/src/orchestrator/block-proving-state.ts

@@ -55,6 +55,7 @@ export class BlockProvingState {
     | ProofState<BlockRollupPublicInputs, typeof NESTED_RECURSIVE_ROLLUP_HONK_PROOF_LENGTH>
     | undefined;
   private builtBlockHeader: BlockHeader | undefined;
+  private builtArchive: AppendOnlyTreeSnapshot | undefined;
   private endState: StateReference | undefined;
   private endSpongeBlob: SpongeBlob | undefined;
   private txs: TxProvingState[] = [];
@@ -232,6 +233,14 @@ export class BlockProvingState {
     return this.builtBlockHeader;
   }
 
+  public setBuiltArchive(archive: AppendOnlyTreeSnapshot) {
+    this.builtArchive = archive;
+  }
+
+  public getBuiltArchive() {
+    return this.builtArchive;
+  }
+
   public getStartSpongeBlob() {
     return this.startSpongeBlob;
   }

package/src/orchestrator/checkpoint-proving-state.ts

@@ -85,7 +85,7 @@ export class CheckpointProvingState {
       typeof L1_TO_L2_MSG_SUBTREE_ROOT_SIBLING_PATH_LENGTH
     >,
     public parentEpoch: EpochProvingState,
-    private onBlobAccumulatorSet: (checkpoint: CheckpointProvingState) => void,
+    private onBlobAccumulatorSet: (checkpoint: CheckpointProvingState) => Promise<void>,
   ) {
     this.blockProofs = new UnbalancedTreeStore(totalNumBlocks);
     this.firstBlockNumber = BlockNumber(headerOfLastBlockInPreviousCheckpoint.globalVariables.blockNumber + 1);
@@ -204,7 +204,7 @@ export class CheckpointProvingState {
       Fr.ZERO,
       NUM_MSGS_PER_BASE_PARITY,
     );
-    return new ParityBasePrivateInputs(messages, this.constants.vkTreeRoot);
+    return new ParityBasePrivateInputs(messages, this.constants.vkTreeRoot, this.constants.proverId);
   }
 
   public setOutHashHint(hint: OutHashHint) {
@@ -245,7 +245,7 @@ export class CheckpointProvingState {
     this.endBlobAccumulator = await accumulateBlobs(this.blobFields!, startBlobAccumulator);
     this.startBlobAccumulator = startBlobAccumulator;
 
-    this.onBlobAccumulatorSet(this);
+    await this.onBlobAccumulatorSet(this);
 
     return this.endBlobAccumulator;
   }
@@ -271,7 +271,7 @@ export class CheckpointProvingState {
     return this.totalNumBlocks === 1 ? 'rollup-checkpoint-root-single-block' : 'rollup-checkpoint-root';
   }
 
-  public getCheckpointRootRollupInputs() {
+  public async getCheckpointRootRollupInputs() {
     const proofs = this.#getChildProofsForRoot();
     const nonEmptyProofs = proofs.filter(p => !!p);
     if (proofs.length !== nonEmptyProofs.length) {
@@ -287,7 +287,7 @@ export class CheckpointProvingState {
     // `blobFields` must've been set if `startBlobAccumulator` is set (in `accumulateBlobs`).
     const blobFields = this.blobFields!;
 
-    const { blobCommitments, blobsHash } = buildBlobHints(blobFields);
+    const { blobCommitments, blobsHash } = await buildBlobHints(blobFields);
 
     const hints = CheckpointRootRollupHints.from({
       previousBlockHeader: this.headerOfLastBlockInPreviousCheckpoint,

package/src/orchestrator/epoch-proving-state.ts

@@ -76,7 +76,7 @@ export class EpochProvingState {
     public readonly epochNumber: EpochNumber,
     public readonly totalNumCheckpoints: number,
     private readonly finalBlobBatchingChallenges: FinalBlobBatchingChallenges,
-    private onCheckpointBlobAccumulatorSet: (checkpoint: CheckpointProvingState) => void,
+    private onCheckpointBlobAccumulatorSet: (checkpoint: CheckpointProvingState) => Promise<void>,
     private completionCallback: (result: ProvingResult) => void,
     private rejectionCallback: (reason: string) => void,
   ) {
@@ -254,9 +254,11 @@ export class EpochProvingState {
       }
       outHashes.push(outHash);
 
-      // Get or create hints for the next checkpoint.
-      hint = checkpoint.getOutHashHintForNextCheckpoint() ?? (await computeOutHashHint(outHashes));
-      checkpoint.setOutHashHintForNextCheckpoint(hint);
+      // If this is NOT the last checkpoint, get or create the hint for the next checkpoint.
+      if (i !== this.totalNumCheckpoints - 1) {
+        hint = checkpoint.getOutHashHintForNextCheckpoint() ?? (await computeOutHashHint(outHashes));
+        checkpoint.setOutHashHintForNextCheckpoint(hint);
+      }
     }
   }
 

package/src/orchestrator/orchestrator.ts

@@ -12,7 +12,9 @@ import { Fr } from '@aztec/foundation/curves/bn254';
 import { AbortError } from '@aztec/foundation/error';
 import { type Logger, type LoggerBindings, createLogger } from '@aztec/foundation/log';
 import { promiseWithResolvers } from '@aztec/foundation/promise';
+import { SerialQueue } from '@aztec/foundation/queue';
 import { assertLength } from '@aztec/foundation/serialize';
+import { sleep } from '@aztec/foundation/sleep';
 import { pushTestData } from '@aztec/foundation/testing';
 import { elapsed } from '@aztec/foundation/timer';
 import type { TreeNodeLocation } from '@aztec/foundation/trees';
@@ -71,11 +73,6 @@ import { EpochProvingState, type ProvingResult, type TreeSnapshots } from './epo
 import { ProvingOrchestratorMetrics } from './orchestrator_metrics.js';
 import { TxProvingState } from './tx-proving-state.js';
 
-type WorldStateFork = {
-  fork: MerkleTreeWriteOperations;
-  cleanupPromise: Promise<void> | undefined;
-};
-
 /**
  * Implements an event driven proving scheduler to build the recursive proof tree. The idea being:
  * 1. Transactions are provided to the scheduler post simulation.
@@ -97,19 +94,22 @@ export class ProvingOrchestrator implements EpochProver {
   private provingPromise: Promise<ProvingResult> | undefined = undefined;
   private metrics: ProvingOrchestratorMetrics;
   // eslint-disable-next-line aztec-custom/no-non-primitive-in-collections
-  private dbs: Map<BlockNumber, WorldStateFork> = new Map();
+  private dbs: Map<BlockNumber, MerkleTreeWriteOperations> = new Map();
   private logger: Logger;
+  private deferredJobQueue = new SerialQueue();
 
   constructor(
     private dbProvider: ReadonlyWorldStateAccess & ForkMerkleTreeOperations,
     private prover: ServerCircuitProver,
     private readonly proverId: EthAddress,
     private readonly cancelJobsOnStop: boolean = false,
+    private readonly enqueueConcurrency: number,
     telemetryClient: TelemetryClient = getTelemetryClient(),
     bindings?: LoggerBindings,
   ) {
     this.logger = createLogger('prover-client:orchestrator', bindings);
     this.metrics = new ProvingOrchestratorMetrics(telemetryClient, 'ProvingOrchestrator');
+    this.deferredJobQueue.start(this.enqueueConcurrency);
   }
 
   get tracer(): Tracer {
@@ -124,9 +124,11 @@ export class ProvingOrchestrator implements EpochProver {
     return this.dbs.size;
   }
 
-  public stop(): Promise<void> {
+  public async stop(): Promise<void> {
+    // Grab the old queue before cancel() replaces it, so we can await its draining.
+    const oldQueue = this.deferredJobQueue;
     this.cancel();
-    return Promise.resolve();
+    await oldQueue.cancel();
   }
 
   public startNewEpoch(
@@ -182,7 +184,7 @@ export class ProvingOrchestrator implements EpochProver {
     const db = await this.dbProvider.fork(lastBlockNumber);
 
     const firstBlockNumber = BlockNumber(lastBlockNumber + 1);
-    this.dbs.set(firstBlockNumber, { fork: db, cleanupPromise: undefined });
+    this.dbs.set(firstBlockNumber, db);
 
     // Get archive sibling path before any block in this checkpoint lands.
     const lastArchiveSiblingPath = await getLastSiblingPath(MerkleTreeId.ARCHIVE, db);
@@ -240,9 +242,9 @@ export class ProvingOrchestrator implements EpochProver {
     if (!this.dbs.has(blockNumber)) {
       // Fork world state at the end of the immediately previous block
       const db = await this.dbProvider.fork(BlockNumber(blockNumber - 1));
-      this.dbs.set(blockNumber, { fork: db, cleanupPromise: undefined });
+      this.dbs.set(blockNumber, db);
     }
-    const db = this.dbs.get(blockNumber)!.fork;
+    const db = this.getDbForBlock(blockNumber);
 
     // Get archive snapshot and sibling path before any txs in this block lands.
     const lastArchiveTreeSnapshot = await getTreeSnapshot(MerkleTreeId.ARCHIVE, db);
@@ -317,7 +319,7 @@ export class ProvingOrchestrator implements EpochProver {
 
     this.logger.info(`Adding ${txs.length} transactions to block ${blockNumber}`);
 
-    const db = this.dbs.get(blockNumber)!.fork;
+    const db = this.getDbForBlock(blockNumber);
     const lastArchive = provingState.lastArchiveTreeSnapshot;
     const newL1ToL2MessageTreeSnapshot = provingState.newL1ToL2MessageTreeSnapshot;
     const spongeBlobState = provingState.getStartSpongeBlob().clone();
@@ -445,14 +447,20 @@ export class ProvingOrchestrator implements EpochProver {
       throw new Error('Block header mismatch');
     }
 
-    // Get db for this block
-    const db = this.dbs.get(provingState.blockNumber)!.fork;
+    // Get db for this block and remove from map — no other code should use it after this point.
+    const db = this.getDbForBlock(provingState.blockNumber);
+    this.dbs.delete(provingState.blockNumber);
 
-    // Update the archive tree, so we're ready to start processing the next block:
-    this.logger.verbose(
-      `Updating archive tree with block ${provingState.blockNumber} header ${(await header.hash()).toString()}`,
-    );
-    await db.updateArchive(header);
+    // Update the archive tree, capture the snapshot, and close the fork deterministically.
+    try {
+      this.logger.verbose(
+        `Updating archive tree with block ${provingState.blockNumber} header ${(await header.hash()).toString()}`,
+      );
+      await db.updateArchive(header);
+      provingState.setBuiltArchive(await getTreeSnapshot(MerkleTreeId.ARCHIVE, db));
+    } finally {
+      await db.close();
+    }
 
     await this.verifyBuiltBlockAgainstSyncedState(provingState);
 
@@ -472,6 +480,13 @@ export class ProvingOrchestrator implements EpochProver {
       this.logger.debug('Block root rollup proof not built yet, skipping header check.');
       return;
     }
+
+    const newArchive = provingState.getBuiltArchive();
+    if (!newArchive) {
+      this.logger.debug('Archive snapshot not yet captured, skipping header check.');
+      return;
+    }
+
     const header = await buildHeaderFromCircuitOutputs(output);
 
     if (!(await header.hash()).equals(await builtBlockHeader.hash())) {
@@ -480,11 +495,7 @@ export class ProvingOrchestrator implements EpochProver {
       return;
     }
 
-    // Get db for this block
     const blockNumber = provingState.blockNumber;
-    const db = this.dbs.get(blockNumber)!.fork;
-
-    const newArchive = await getTreeSnapshot(MerkleTreeId.ARCHIVE, db);
     const syncedArchive = await getTreeSnapshot(MerkleTreeId.ARCHIVE, this.dbProvider.getSnapshot(blockNumber));
     if (!syncedArchive.equals(newArchive)) {
       this.logger.error(
@@ -502,12 +513,6 @@ export class ProvingOrchestrator implements EpochProver {
       provingState.reject(`New archive mismatch.`);
       return;
     }
-
-    // TODO(palla/prover): This closes the fork only on the happy path. If this epoch orchestrator
-    // is aborted and never reaches this point, it will leak the fork. We need to add a global cleanup,
-    // but have to make sure it only runs once all operations are completed, otherwise some function here
-    // will attempt to access the fork after it was closed.
-    void this.cleanupDBFork(blockNumber);
   }
 
   /**
@@ -516,6 +521,11 @@ export class ProvingOrchestrator implements EpochProver {
    * If cancelJobsOnStop is false (default), jobs remain in the broker queue and can be reused on restart/reorg.
    */
   public cancel() {
+    void this.deferredJobQueue.cancel();
+    // Recreate the queue so it can accept jobs for subsequent epochs.
+    this.deferredJobQueue = new SerialQueue();
+    this.deferredJobQueue.start(this.enqueueConcurrency);
+
     if (this.cancelJobsOnStop) {
       for (const controller of this.pendingProvingJobs) {
         controller.abort();
@@ -523,6 +533,19 @@ export class ProvingOrchestrator implements EpochProver {
     }
 
     this.provingState?.cancel();
+
+    for (const [blockNumber, db] of this.dbs.entries()) {
+      void db.close().catch(err => this.logger.error(`Error closing db for block ${blockNumber}`, err));
+    }
+    this.dbs.clear();
+  }
+
+  private getDbForBlock(blockNumber: BlockNumber): MerkleTreeWriteOperations {
+    const db = this.dbs.get(blockNumber);
+    if (!db) {
+      throw new Error(`World state fork for block ${blockNumber} not found.`);
+    }
+    return db;
   }
 
   /**
@@ -554,24 +577,6 @@ export class ProvingOrchestrator implements EpochProver {
     return epochProofResult;
   }
 
-  private async cleanupDBFork(blockNumber: BlockNumber): Promise<void> {
-    this.logger.debug(`Cleaning up world state fork for ${blockNumber}`);
-    const fork = this.dbs.get(blockNumber);
-    if (!fork) {
-      return;
-    }
-
-    try {
-      if (!fork.cleanupPromise) {
-        fork.cleanupPromise = fork.fork.close();
-      }
-      await fork.cleanupPromise;
-      this.dbs.delete(blockNumber);
-    } catch (err) {
-      this.logger.error(`Error closing db for block ${blockNumber}`, err);
-    }
-  }
-
   /**
    * Enqueue a job to be scheduled
    * @param provingState - The proving state object being operated on
@@ -630,8 +635,11 @@ export class ProvingOrchestrator implements EpochProver {
       }
     };
 
-    // let the callstack unwind before adding the job to the queue
-    setImmediate(() => void safeJob());
+    void this.deferredJobQueue.put(async () => {
+      void safeJob();
+      // we yield here to the macro task queue such to give Nodejs a chance to run other operatoins in between enqueues
+      await sleep(0);
+    });
   }
 
   private async updateL1ToL2MessageTree(l1ToL2Messages: Fr[], db: MerkleTreeWriteOperations) {
@@ -894,17 +902,15 @@ export class ProvingOrchestrator implements EpochProver {
         const leafLocation = provingState.setBlockRootRollupProof(result);
         const checkpointProvingState = provingState.parentCheckpoint;
 
-        // If the proofs were slower than the block header building, then we need to try validating the block header hashes here.
+        // Verification is called from both here and setBlockCompleted. Whichever runs last
+        // will be the first to see all three pieces (header, proof output, archive) and run the checks.
         await this.verifyBuiltBlockAgainstSyncedState(provingState);
 
         if (checkpointProvingState.totalNumBlocks === 1) {
-          this.checkAndEnqueueCheckpointRootRollup(checkpointProvingState);
+          await this.checkAndEnqueueCheckpointRootRollup(checkpointProvingState);
         } else {
-          this.checkAndEnqueueNextBlockMergeRollup(checkpointProvingState, leafLocation);
+          await this.checkAndEnqueueNextBlockMergeRollup(checkpointProvingState, leafLocation);
         }
-
-        // We are finished with the block at this point, ensure the fork is cleaned up
-        void this.cleanupDBFork(provingState.blockNumber);
       },
     );
   }
@@ -1009,14 +1015,14 @@ export class ProvingOrchestrator implements EpochProver {
         },
         signal => this.prover.getBlockMergeRollupProof(inputs, signal, provingState.epochNumber),
       ),
-      result => {
+      async result => {
         provingState.setBlockMergeRollupProof(location, result);
-        this.checkAndEnqueueNextBlockMergeRollup(provingState, location);
+        await this.checkAndEnqueueNextBlockMergeRollup(provingState, location);
       },
     );
   }
 
-  private enqueueCheckpointRootRollup(provingState: CheckpointProvingState) {
+  private async enqueueCheckpointRootRollup(provingState: CheckpointProvingState) {
     if (!provingState.verifyState()) {
       this.logger.debug('Not running checkpoint root rollup. State no longer valid.');
       return;
@@ -1031,7 +1037,7 @@ export class ProvingOrchestrator implements EpochProver {
 
     this.logger.debug(`Enqueuing ${rollupType} for checkpoint ${provingState.index}.`);
 
-    const inputs = provingState.getCheckpointRootRollupInputs();
+    const inputs = await provingState.getCheckpointRootRollupInputs();
 
     this.deferredProving(
       provingState,
@@ -1191,25 +1197,28 @@ export class ProvingOrchestrator implements EpochProver {
     this.enqueueBlockRootRollup(provingState);
   }
 
-  private checkAndEnqueueNextBlockMergeRollup(provingState: CheckpointProvingState, currentLocation: TreeNodeLocation) {
+  private async checkAndEnqueueNextBlockMergeRollup(
+    provingState: CheckpointProvingState,
+    currentLocation: TreeNodeLocation,
+  ) {
     if (!provingState.isReadyForBlockMerge(currentLocation)) {
       return;
     }
 
     const parentLocation = provingState.getParentLocation(currentLocation);
     if (parentLocation.level === 0) {
-      this.checkAndEnqueueCheckpointRootRollup(provingState);
+      await this.checkAndEnqueueCheckpointRootRollup(provingState);
     } else {
       this.enqueueBlockMergeRollup(provingState, parentLocation);
     }
   }
 
-  private checkAndEnqueueCheckpointRootRollup(provingState: CheckpointProvingState) {
+  private async checkAndEnqueueCheckpointRootRollup(provingState: CheckpointProvingState) {
     if (!provingState.isReadyForCheckpointRoot()) {
       return;
     }
 
-    this.enqueueCheckpointRootRollup(provingState);
+    await this.enqueueCheckpointRootRollup(provingState);
   }
 
   private checkAndEnqueueNextCheckpointMergeRollup(provingState: EpochProvingState, currentLocation: TreeNodeLocation) {

package/src/prover-client/prover-client.ts

@@ -54,6 +54,7 @@ export class ProverClient implements EpochProverManager {
       facade,
       this.config.proverId,
       this.config.cancelJobsOnStop,
+      this.config.enqueueConcurrency,
       this.telemetry,
       bindings,
     );
@@ -156,7 +157,7 @@ export class ProverClient implements EpochProverManager {
 }
 
 export function buildServerCircuitProver(
-  config: ActualProverConfig & ACVMConfig & BBConfig,
+  config: Omit<ActualProverConfig, 'enqueueConcurrency'> & ACVMConfig & BBConfig,
   telemetry: TelemetryClient,
 ): Promise<ServerCircuitProver> {
   if (config.realProofs) {

package/src/proving_broker/broker_prover_facade.ts

@@ -4,8 +4,9 @@ import type {
   NESTED_RECURSIVE_ROLLUP_HONK_PROOF_LENGTH,
   RECURSIVE_PROOF_LENGTH,
 } from '@aztec/constants';
+import { asyncPool } from '@aztec/foundation/async-pool';
 import { EpochNumber } from '@aztec/foundation/branded-types';
-import { sha256 } from '@aztec/foundation/crypto/sha256';
+import { chunk } from '@aztec/foundation/collection';
 import { type Logger, type LoggerBindings, createLogger } from '@aztec/foundation/log';
 import { type PromiseWithResolvers, RunningPromise, promiseWithResolvers } from '@aztec/foundation/promise';
 import { truncate } from '@aztec/foundation/string';
@@ -46,6 +47,8 @@ import type {
   TxRollupPublicInputs,
 } from '@aztec/stdlib/rollup';
 
+import { createHash } from 'node:crypto';
+
 import { InlineProofStore, type ProofStore } from './proof_store/index.js';
 
 // Perform a snapshot sync every 30 seconds
@@ -225,17 +228,11 @@ export class BrokerCircuitProverFacade implements ServerCircuitProver {
       // We collect all returned notifications and return them
       const allCompleted = new Set<ProvingJobId>();
       try {
-        let numRequests = 0;
-        while (ids.length > 0) {
-          const slice = ids.splice(0, SNAPSHOT_SYNC_CHECK_MAX_REQUEST_SIZE);
-          const completed = await this.broker.getCompletedJobs(slice);
+        const batches = ids.length > 0 ? chunk(ids, SNAPSHOT_SYNC_CHECK_MAX_REQUEST_SIZE) : [[]];
+        await asyncPool(1, batches, async batch => {
+          const completed = await this.broker.getCompletedJobs(batch);
           completed.forEach(id => allCompleted.add(id));
-          ++numRequests;
-        }
-        if (numRequests === 0) {
-          const final = await this.broker.getCompletedJobs([]);
-          final.forEach(id => allCompleted.add(id));
-        }
+        });
       } catch (err) {
         this.log.error(`Error thrown when requesting completed job notifications from the broker`, err);
       }
@@ -351,12 +348,8 @@ export class BrokerCircuitProverFacade implements ServerCircuitProver {
       .map(id => this.jobs.get(id)!)
       .filter(x => x !== undefined);
     const totalJobsToRetrieve = toBeRetrieved.length;
-    let totalJobsRetrieved = 0;
-    while (toBeRetrieved.length > 0) {
-      const slice = toBeRetrieved.splice(0, MAX_CONCURRENT_JOB_SETTLED_REQUESTS);
-      const results = await Promise.all(slice.map(job => processJob(job!)));
-      totalJobsRetrieved += results.filter(x => x).length;
-    }
+    const results = await asyncPool(MAX_CONCURRENT_JOB_SETTLED_REQUESTS, toBeRetrieved, job => processJob(job));
+    const totalJobsRetrieved = results.filter(x => x).length;
     if (totalJobsToRetrieve > 0) {
       this.log.verbose(
         `Successfully retrieved ${totalJobsRetrieved} of ${totalJobsToRetrieve} jobs that should be ready, total ready jobs is now: ${this.jobsToRetrieve.size}`,
@@ -659,8 +652,12 @@ export class BrokerCircuitProverFacade implements ServerCircuitProver {
     );
   }
 
-  private generateId(type: ProvingRequestType, inputs: { toBuffer(): Buffer }, epochNumber = EpochNumber.ZERO) {
-    const inputsHash = sha256(inputs.toBuffer());
-    return makeProvingJobId(epochNumber, type, inputsHash.toString('hex'));
+  private generateId(
+    type: ProvingRequestType,
+    inputs: { toBuffer(): Buffer },
+    epochNumber = EpochNumber.ZERO,
+  ): ProvingJobId {
+    const inputsHash = createHash('sha256').update(inputs.toBuffer()).digest('hex');
+    return makeProvingJobId(epochNumber, type, inputsHash);
   }
 }

package/src/proving_broker/config.ts

@@ -4,10 +4,11 @@ import {
   booleanConfigHelper,
   getDefaultConfig,
   numberConfigHelper,
+  optionalNumberConfigHelper,
 } from '@aztec/foundation/config';
 import { pickConfigMappings } from '@aztec/foundation/config';
-import { type DataStoreConfig, dataConfigMappings } from '@aztec/kv-store/config';
 import { type ChainConfig, chainConfigMappings } from '@aztec/stdlib/config';
+import { type DataStoreConfig, dataConfigMappings } from '@aztec/stdlib/kv-store';
 import { ProvingRequestType } from '@aztec/stdlib/proofs';
 
 import { z } from 'zod';
@@ -73,7 +74,7 @@ export const proverBrokerConfigMappings: ConfigMappingsType<ProverBrokerConfig>
   },
   proverBrokerStoreMapSizeKb: {
     env: 'PROVER_BROKER_STORE_MAP_SIZE_KB',
-    parseEnv: (val: string | undefined) => (val ? +val : undefined),
+    ...optionalNumberConfigHelper(),
     description: "The size of the prover broker's database. Will override the dataStoreMapSizeKb if set.",
   },
   proverBrokerDebugReplayEnabled: {

package/src/proving_broker/proving_broker.ts

@@ -314,13 +314,17 @@ export class ProvingBroker implements ProvingJobProducer, ProvingJobConsumer, Pr
     // notify listeners of the cancellation
     if (!this.resultsCache.has(id)) {
       this.logger.info(`Cancelling job id=${id}`, { provingJobId: id });
-      await this.#reportProvingJobError(id, 'Aborted', false);
+      await this.#reportProvingJobError(id, 'Aborted', false, undefined, true);
     }
   }
 
   private cleanUpProvingJobState(ids: ProvingJobId[]) {
     for (const id of ids) {
       this.jobsCache.delete(id);
+      const deferred = this.promises.get(id);
+      if (deferred) {
+        deferred.resolve({ status: 'rejected', reason: 'Proving job cleaned up' });
+      }
       this.promises.delete(id);
       this.resultsCache.delete(id);
       this.inProgress.delete(id);
@@ -395,6 +399,7 @@ export class ProvingBroker implements ProvingJobProducer, ProvingJobConsumer, Pr
     err: string,
     retry = false,
     filter?: ProvingJobFilter,
+    aborted = false,
   ): Promise<GetProvingJobResponse | undefined> {
     const info = this.inProgress.get(id);
     const item = this.jobsCache.get(id);
@@ -455,7 +460,11 @@ export class ProvingBroker implements ProvingJobProducer, ProvingJobConsumer, Pr
     this.promises.get(id)!.resolve(result);
     this.completedJobNotifications.push(id);
 
-    this.instrumentation.incRejectedJobs(item.type);
+    if (aborted) {
+      this.instrumentation.incAbortedJobs(item.type);
+    } else {
+      this.instrumentation.incRejectedJobs(item.type);
+    }
     if (info) {
       const duration = this.msTimeSource() - info.startedAt;
       this.instrumentation.recordJobDuration(item.type, duration);
@@ -603,7 +612,7 @@ export class ProvingBroker implements ProvingJobProducer, ProvingJobConsumer, Pr
     const jobsToClean: ProvingJobId[] = [];
     for (const id of jobIds) {
       const job = this.jobsCache.get(id)!;
-      if (this.isJobStale(job)) {
+      if (this.isJobStale(job) && !this.inProgress.has(id) && !this.resultsCache.has(id)) {
         jobsToClean.push(id);
       }
     }
@@ -627,10 +636,26 @@ export class ProvingBroker implements ProvingJobProducer, ProvingJobConsumer, Pr
       const now = this.msTimeSource();
       const msSinceLastUpdate = now - metadata.lastUpdatedAt;
       if (msSinceLastUpdate >= this.jobTimeoutMs) {
-        this.logger.warn(`Proving job id=${id} timed out. Adding it back to the queue.`, { provingJobId: id });
         this.inProgress.delete(id);
-        this.enqueueJobInternal(item);
         this.instrumentation.incTimedOutJobs(item.type);
+
+        const retries = this.retries.get(id) ?? 0;
+        if (retries + 1 < this.maxRetries && !this.isJobStale(item)) {
+          this.logger.warn(`Proving job id=${id} timed out. Re-enqueueing (retry ${retries + 1}/${this.maxRetries}).`, {
+            provingJobId: id,
+          });
+          this.retries.set(id, retries + 1);
+          this.enqueueJobInternal(item);
+        } else {
+          this.logger.error(`Proving job id=${id} timed out after ${retries + 1} attempts. Marking as failed.`, {
+            provingJobId: id,
+          });
+          const result: ProvingJobSettledResult = { status: 'rejected', reason: 'Timed out' };
+          this.resultsCache.set(id, result);
+          this.promises.get(id)?.resolve(result);
+          this.completedJobNotifications.push(id);
+          this.instrumentation.incRejectedJobs(item.type);
+        }
       }
     }
   }