Hitting a persistent build-scheduling failure on a Node/pnpm service (Nixpacks). Same error for over an hour now, on every push and every manual redeploy.

### What we see

Every attempt follows the same shape:

1. Snapshot received ✓

2. Nixpacks analyses, plan prints correctly ✓

3. Saved output to: snapshot-target-unpack ✓

4. scheduling build on Metal builder "builder-dtxatp" ✓

5. ~15 seconds later: failure

Error line, identical every time:

```

Build Failed: build daemon returned an error

< failed to receive status: rpc error: code = NotFound desc = no such job - >

```

None of the install, build or start commands ever execute.

### Failed job IDs (UTC)

| # | Time | Job ID | Trigger |

|---|-------|-----------------------------------------------------------|-----------------------|

| 1 | 09:21 | 7385a945-87b8-4280-886e-00383ca373b6-20260423092134 | git push |

| 2 | 09:28 | 1bbe2163-7e4c-4c66-972b-ee69c22d02f3-20260423092847 | manual Redeploy |

| 3 | 09:33 | 3dfc04ca-c513-47dc-be84-81237bf1e19f-20260423093309 | empty-commit re-push |

| 4 | ~10:30 | (identical error and pattern) | manual Redeploy |

Attempts (1)(2)(3) all explicitly log builder-dtxatp. Attempt (4) log shape identical.

### What we've ruled out

Between the last successful deploy and HEAD (22 commits), none of these files changed:

- railway.json, Dockerfile, nixpacks.toml, .dockerignore

- pnpm-lock.yaml (last changed >1 month ago)

- root package.json, apps/api/package.json, workspace package.jsons

- any tsconfig.json

The Nixpacks plan that prints during the failed analyses is byte-identical to the plan printed on prior successful deploys (same setup / install / build / start). Snapshot size is ~8.3 MB, unchanged shape.

The 22 commits touched TypeScript app code, tests (not in the production build), standalone scripts (not bundled), and markdown docs.

### What we've tried

1. Manual redeploy from the dashboard — same error

2. Empty commit push to invalidate any cached snapshot — same error (new snapshot hash, same outcome)

3. Waited >1hr for any transient builder issue to recover — still same error

4. Checked status.railway.com — reports fully operational