fix(executor): subflow edge keys mismatch

[icecrasher321]

Summary

Test this workflow. Consecutive loops case with empty followed by populated breaks due to edge key mismatch.

Manual trigger
↓
[Seed] (Function)
return { empty: [], populated: [1, 2, 3] }
↓
[LoopEmpty] forEach <seed.result.empty>
└─ body: Function return <loop.currentItem>
↓ (loop-end-source → loop container)
[LoopWork] forEach <seed.result.populated>
└─ body: Function return <loop.currentItem>

Type of Change

• Bug fix

Testing

Tested pre and post fix. Added unit tests.

Checklist

• Code follows project style guidelines
• Self-reviewed my changes
• Tests added/updated and passing
• No new warnings introduced
• I confirm that I have read and agree to the terms outlined in the Contributor License Agreement (CLA)

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
docs	Ready	Preview, Comment	Apr 16, 2026 8:39pm

[cursor]

PR Summary

Medium Risk
Touches core DAG edge-activation/reset logic and subflow container logging, which can affect execution ordering and loop/parallel readiness; changes are well-covered by new regressions but could impact complex workflows.

Overview
Fixes a loop-reset regression where EdgeManager.clearDeactivatedEdgesForNodes could incorrectly re-activate externally cascade-deactivated edges due to substring matching, causing subsequent loop iterations (and consecutive loops) to stall with phantom incoming dependencies.

Adds targeted regression tests around loop sentinel start/end readiness and internal-vs-external edge reset behavior, and refactors loop/parallel orchestrators to emit a new emitSubflowSuccessEvents BlockLog/onComplete callback on successful container completion so trace-span grouping can use the user-configured container names (with new trace-span tests covering this).

^{Reviewed by Cursor Bugbot for commit 5e143b6. Configure here.}

[greptile-apps]

Greptile Summary

This PR fixes a stall bug where consecutive loops — an empty one followed by a populated one — would cause the second loop to hang after its first iteration. The root cause was clearDeactivatedEdgesForNodes using includes(-${nodeId}-) which inadvertently matched edge keys where the loop's sentinel was the target (not the source), re-activating cascade-deactivated external edges and leaving a phantom pending incoming that blocked isNodeReady. The fix narrows the check to startsWith(${nodeId}-) so only edges originating inside the loop are cleared on reset. As a secondary change, a new emitSubflowSuccessEvents utility deduplicates the onBlockComplete/BlockLog emission for successful containers across both loop and parallel orchestrators, and ensures the user-configured block name propagates to trace spans.

Confidence Score: 5/5

Safe to merge — the one-line edge-key fix is well-reasoned, backed by three targeted regression tests, and the remaining finding is a P2 style cleanup.

All substantive findings are P2 or lower (unused return type). The core bug fix is correct, well-documented, and fully tested. The utility refactor cleanly deduplicates loop/parallel completion logic without changing observable behavior.

apps/sim/executor/utils/subflow-utils.ts — minor unused Promise return type in emitSubflowSuccessEvents

Important Files Changed

Filename	Overview
apps/sim/executor/execution/edge-manager.ts	Core bug fix: narrows clearDeactivatedEdgesForNodes to only remove edges whose SOURCE is inside the loop, preventing external cascade-deactivated edges from being incorrectly re-activated during loop reset.
apps/sim/executor/execution/edge-manager.test.ts	Three new regression tests covering the fixed bug, the post-reset loop-continue path, and a guard ensuring internal loop edges are still cleared — thorough coverage for the edge-key matching change.
apps/sim/executor/utils/subflow-utils.ts	New emitSubflowSuccessEvents utility deduplicates the onBlockComplete/BlockLog logic for successful containers; minor issue with an unused Promise return type.
apps/sim/executor/orchestrators/loop.ts	Clean refactor: inlined onBlockComplete block replaced by emitSubflowSuccessEvents call, removing ~25 lines of duplicated logic.
apps/sim/executor/orchestrators/parallel.ts	Same refactor as loop.ts — inlined onBlockComplete replaced by emitSubflowSuccessEvents, consistent with loop orchestrator.
apps/sim/lib/logs/execution/trace-spans/trace-spans.ts	Comment updated to reflect that container BlockLogs are now emitted on success (not just skip/error), enabling resolveContainerName to use the user-configured name.
apps/sim/lib/logs/execution/trace-spans/trace-spans.test.ts	Two new concurrent tests verifying that the user-configured loop/parallel container name appears in the trace span when a success BlockLog is present.

Sequence Diagram

sequenceDiagram
    participant External as External Node
    participant SentinelStart as Loop Sentinel-Start
    participant Body as Loop Body
    participant SentinelEnd as Loop Sentinel-End
    participant EdgeMgr as EdgeManager

    Note over External,EdgeMgr: Empty upstream loop cascades deactivation
    External->>EdgeMgr: processOutgoingEdges({selectedOption: 'else'})
    EdgeMgr->>EdgeMgr: deactivateEdgeAndDescendants(external-node, sentinel-start)
    Note over EdgeMgr: deactivatedEdges = {external-node-sentinel-start-...}

    Note over External,EdgeMgr: Populated loop resets for next iteration
    EdgeMgr->>EdgeMgr: clearDeactivatedEdgesForNodes({sentinel-start, body, sentinel-end})
    Note over EdgeMgr: OLD: includes(-sentinel-start-) also matched<br/>external-node-sentinel-start-... → wrongly cleared!
    Note over EdgeMgr: NEW: startsWith(sentinel-start-) only matches<br/>edges where sentinel-start is SOURCE → external edge preserved

    SentinelEnd->>EdgeMgr: processOutgoingEdges({selectedRoute: loop_continue})
    EdgeMgr->>SentinelStart: incomingEdges: only loop_continue active
    Note over SentinelStart: countActiveIncomingEdges = 0 → READY ✓

Loading

_{Reviews (1): Last reviewed commit: "fix(executor): subflow edge keys mismatc..." | Re-trigger Greptile}

[icecrasher321]

bugbot run

[cursor]

✅ Bugbot reviewed your changes and found no new issues!

Comment @cursor review or bugbot run to trigger another review on this PR

^{Reviewed by Cursor Bugbot for commit 5e143b6. Configure here.}