Cancelled video-to-software request leaks the scaffold temp directory

[groupthinking]

Problem

Once ProjectCodeGenerator moved its scaffolding filesystem work off the event loop (#1250 / #1251), the off-loop hops introduced await (cancellation) points that did not exist when the writes ran inline. A POST /api/v1/video-to-software request that is cancelled or fails mid-scaffold can now leave a uvai_project_* temporary directory that no caller receives or removes.

asyncio.to_thread cannot interrupt a worker thread once it has started, so a cancellation delivered while a scaffolding hop runs raises CancelledError in the coroutine while the thread keeps writing to disk. generate_project never returns the directory path, so nothing cleans it up.

There are two leak windows:

1. mkdtemp window — cancellation delivered while tempfile.mkdtemp runs; the directory is created but its path is never assigned into a cleanup scope.
2. write-plan window — cancellation delivered while _apply_write_plan runs; the scaffold is partially/fully written and abandoned.

Acceptance criteria

• A request cancelled or failed at any point after the scaffold directory is created removes that directory before propagating.
• Off-loop hops wait for their worker to settle before propagating cancellation, so cleanup never races a live writer.
• Regression tests cover both windows (cancel during mkdtemp, cancel during _apply_write_plan, and failure during the write plan).
• Happy-path generated output is unchanged (byte-for-byte equality preserved from perf: batch project scaffolding disk writes off the event loop #1251).

Fix

Tracked in PR #1252 on branch claude/determined-maxwell-9t5kou.

[linear-code]

GRV-252

[github-actions]

👋 Thanks for opening this issue! It has been automatically labeled for triage.

[coderabbitai]

🔗 Related PRs

#313 - fix(pipeline): stream API key on job poll, async jobs, smoke SSE path [merged]
#314 - fix(pipeline): async kickoff uses videos/process [merged]

---

📝 Issue Planner

_{Check the box below or use the @coderabbitai plan command to generate an implementation plan and prompts that you can use with your favorite coding assistant.}

• Create Plan

---

🧪 Issue enrichment is currently in open beta.

You can configure auto-planning by selecting labels in the issue_enrichment configuration.

To disable automatic issue enrichment, add the following to your .coderabbit.yaml:

issue_enrichment:
  auto_enrich:
    enabled: false

💬 Have feedback or questions? Drop into our discord!