fix(mount): classify transport rejections from the launch and daemon-start execs

[khaliqgant]

Refs #45.

The gap

startMount wraps its mkdir and status-probe execs so a rejected transport becomes a classified error, but leaves two unwrapped:

exec	wrapped before	on rejection
mkdir mount path	yes	classified
initial-sync launch	no	unclassifiable
status probe (poll loop)	yes	classified
daemon start	no	unclassifiable

Every throw in this function that embeds sandbox command output is one callers recognise by message prefix. So those two were the only paths that could produce an error matching no prefix at all — and callers bucket that as "invocation unknown" and drop the cause.

Why it matters right now

Cloud is failing 100% of JIT sandbox provisions on exactly that bucket. Measured 5/5 against production:

5/5  HTTP 503  phase: relayfile_mount_invocation_unknown
     mountElapsedMs 112143 / 131990 / 122960 / 133527 / 117733

~122s matches the ~120s Daytona proxy read timeout this file's own comment cites. The underlying error was unrecoverable from either the logs or the HTTP body, which is how this got chased as a Daytona provisioning-timeout problem for a day while Daytona was in fact creating the sandbox in 1.3 seconds.

Change

Wrap both in the same shape the two already-wrapped execs use, preserving the original as cause. A rejection now lands on Failed to launch relayfile initial sync: or Failed to start relayfile mount:, which callers already map to named phases (relayfile_mount_daemon_invoked_initial_sync_not_invoked and relayfile_mount_invoked_exited).

This makes the failure legible; it does not stop the exec from hanging. #45 stays open for the hang itself. Paired with AgentWorkforce/cloud#3244 (merged), the next occurrence will name which of the two died instead of returning an opaque phase — which is what the hang investigation needs to start from evidence rather than inference.

Tests

npm test — 819 pass, 0 fail. npm run typecheck clean.

Two new tests in orchestrator.start-mount.test.ts, mirroring the existing mkdir/status rejection tests: each asserts the classified message and that error.cause is the original transport error. The daemon-start one drives the full happy path first, since that exec is only reached after the initial sync reports a clean exit.

Red-check — reverting both wraps fails exactly the two new tests (# pass 817 # fail 2); restored, 819/0.

🤖 Generated with Claude Code

https://claude.ai/code/session_01AGB949mfHcP68whBJebzEM

---

Summary by cubic

Fixes startMount so transport rejections from the initial-sync launch and daemon-start execs are no longer unclassifiable. Previously only the mkdir and status-probe execs were wrapped, so failures in these two paths matched no caller-recognized message prefix, landed in the "invocation unknown" bucket, and lost the original cause.

Related to #45. Cloud is failing all JIT sandbox provisions on that unknown bucket after ~122s, which matches the ~120s Daytona proxy read timeout, and the underlying error was unrecoverable from logs or the HTTP body.

• Wraps both execs in the same classified shape the existing ones use, preserving the original transport error as cause.
• Rejections now surface as Failed to launch relayfile initial sync: or Failed to start relayfile mount:, which callers already map to named phases.
• Makes the failure legible but does not stop the exec from hanging; startMount: two unwrapped execs make a transport rejection unclassifiable (100% JIT provision failure in prod) #45 stays open for the hang itself.

^{Written for commit 7765f12. Summary will update on new commits.}

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-01T13:05:27.003298Z	7765f12	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Team

Run ID: 6618c9a3-b590-47a8-bdef-52768e8e4663

📥 Commits

Reviewing files that changed from the base of the PR and between d68a71b and 7765f12.

📒 Files selected for processing (2)

• src/orchestrator.start-mount.test.ts
• src/orchestrator.ts

Included review availability: Your plan provides up to 1 included review per hour; 0 remain after this review.

---

📝 Walkthrough

Walkthrough

startMount now adds context to initial-sync and mount-start failures. Each wrapped error preserves the original adapter error as cause. Tests cover both failure points and command execution order.

Changes

startMount error handling

Layer / File(s)	Summary
Contextual error wrapping and validation src/orchestrator.ts, src/orchestrator.start-mount.test.ts	startMount wraps initial-sync and mount-start failures with operation-specific messages and preserves the original errors as cause. Tests verify both classifications and expected command execution.

Estimated code review effort: 2 (Simple) | ~10 minutes

Merge Risk: ⚪ Minimal · up to 7765f

The change adds operation-specific context to existing mount execution failures without altering mount behavior or system boundaries. It is merge-ready after normal checks and review, with no actionable merge-blocking risk remaining.

Suggested reviewers: miyaontherelay, kjgbot

Poem

A rabbit hops where start scripts run
Errors now tell what went wrong
Their causes stay tucked safely near
Tests check the path from far to here
The relay mount begins with cheer

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly and concisely identifies the main change: classification of transport rejections from the launch and daemon-start executions.
Description check	✅ Passed	The description directly explains the unwrapped execution paths, the new error classification, cause preservation, tests, and the scope limit regarding execution hangs.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 2…
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

Full details: Docstring Coverage

Explanation

No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 2 files.

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch fix/classify-mount-exec-rejections

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[cubic-dev-ai]

No issues found across 2 files

_{Re-trigger cubic}