feat: use GenerationID to gate stale checkpoints

[Linell]

Summary

Echoes generation_id from the executor on every async checkpoint POST and treats a 409 response as a stale dispatch (StaleDispatchError), halting execution instead of returning buffered ops. Without this, an HTTP timeout that causes the executor to requeue can race with a late checkpoint: the executor would memoize the returned ops as canonical and chain the next dispatch off a dead invocation, producing duplicate step executions. Pairs with inngest/inngest#4108.

Checklist

• Added a docs PR that references this PR
• Added unit/integration tests
• Added changesets if applicable

Related

EXE-1552: HTTP timeout leads to duplicate steps with checkpointing
inngest/inngest#4108
inngest/inngestgo#213

---

Note

Echoes generation_id from the executor on every async checkpoint POST and treats a 409 response as a StaleDispatchError, halting execution instead of returning buffered ops. This prevents a race condition where an HTTP timeout causes the executor to requeue while a late checkpoint returns buffered ops that get memoized as canonical, producing duplicate step executions.

^{Written by Mendral for commit 7174549.}

[linear-code]

EXE-1552 HTTP timeout leads to duplicate steps with checkpointing

[changeset-bot]

🦋 Changeset detected

Latest commit: 7174549

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 1 package

Name	Type
inngest	Minor

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[mendral-app]

Needs attention — 1 issue in 1 file

The core logic is sound — the 409 detection, shouldRetry short-circuit, and retriable: false halt are all correct. One bug: the falsy guard on generationId silently drops the field when the value is 0, which is a valid generation ID.

Prompt for AI agents (all issues)

Check if these issues are valid — if so, understand the root cause of each and fix them.

<assessment>
The core logic is sound — the 409 detection, `shouldRetry` short-circuit, and `retriable: false` halt are all correct. One bug: the falsy guard on `generationId` silently drops the field when the value is `0`, which is a valid generation ID.
</assessment>

<file name="packages/inngest/src/api/api.ts">
<issue location="packages/inngest/src/api/api.ts:560">
`generationId: 0` is silently dropped because `0` is falsy. If the executor ever issues generation ID 0, the checkpoint POST omits `generation_id` entirely and the 409 guard never fires.
</issue>
</file>

_{Tag @mendral-app with feedback or questions. View session}

[mendral-app]

bug (P1): generationId: 0 is silently dropped because 0 is falsy. If the executor ever issues generation ID 0, the checkpoint POST omits generation_id entirely and the 409 guard never fires.

Suggested change

Suggested change

	...(args.generationId ? { generation_id: args.generationId } : {}),
	...(args.generationId !== undefined ? { generation_id: args.generationId } : {}),

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At packages/inngest/src/api/api.ts, line 560:

<issue>
`generationId: 0` is silently dropped because `0` is falsy. If the executor ever issues generation ID 0, the checkpoint POST omits `generation_id` entirely and the 409 guard never fires.
</issue>