Build fails with "railpack prepare exited with an error" (no detail) — reproducible on a previously-successful commit

Question

Summary

Every deploy for two services in one project has failed since ~2026-08-29 10:00 UTC with a single terse build-log line and no further diagnostic detail: `railpack prepare exited with an error`. The last successful deploy for both services was 2026-08-28 19:16:24 UTC — nothing in either service's config changed between then and now. I isolated this to a platform-side regression, not our code: redeploying the exact commit that succeeded 15 hours earlier, through the identical path, fails identically.

Environment

Every failed attempt produced byte-identical build log output (only the snapshot hash/size and builder-node name differ).

Isolation testing performed

1. CLI vs API upload path — reproduced via both `railway up --ci` and the deploy API. Same failure both ways.
2. Builder node — deploys were scheduled on two different Metal builder nodes (builder-cnronl and builder-szkett observed in logs). Same failure on both.
3. Service-specific config/cache — reproduced on two different services in the same project with identical config. Same failure on both.
4. Our code — deployed the exact commit (9a7cf52) that produced the last successful build (2026-08-28 19:16:24 UTC), through the identical deploy path. It failed identically (deployment 62f290a5-4a99-442d-b903-59467f3a999f). I also bisected our new commit's diff (deploying various subsets of the changed files) — every variant failed the same way, including zero code changes at all. This rules out our code as the cause.
5. Platform status — status.railway.com showed "Fully Operational" at the time, no incident listed (acknowledged that page only tracks broad/high-impact incidents). Separately, browsing Central Station's own "Newest Questions" right now, I see several other users reporting build/deploy failures in this same window (Aug 28 evening onward), including one titled "The status for the most recent incident say resolved, but my deployments are still failing" — which suggests this may not be isolated to our project.

Full build log (representative — every attempt above matches this)

scheduling build on Metal builder "builder-cnronl"

fetched snapshot sha256:60e7903592dbebba5ce0deab64d8c49e9f285a51b677001fac45eef6ea47fbdf (585 kB bytes)

fetching snapshot

unpacking archive

using build driver railpack-v0.38.0

railpack prepare exited with an error

No further stdout/stderr is exposed — this is the complete log, confirmed via both the dashboard and `railway logs --build --json`.

Impact

Production is not down — both services are still serving their last successful build, verified live (/healthz returns healthy, application endpoints return 200). We simply cannot ship new deploys until this is resolved.

Ask

Since our own testing rules out our code, config, and a single bad builder node, we suspect a regression in railpack-v0.38.0's "prepare" step itself, or something in our project's cached build state on Railway's side. Could you:

1. Surface the actual underlying error from railpack's prepare step (the platform-side log only shows the one-line wrapper above, no cause).
2. Check whether anything changed in this project's build pipeline / cached build state between 2026-08-28 19:16 UTC and 2026-08-29 10:00 UTC.

Happy to provide any additional deployment IDs, account info, or run further tests on request.