When the remote CAS (compilation) cache becomes unavailable, for example during a backend pod roll, a lost box, or a network blip, builds no longer stall waiting on per-request timeouts. Tuist now detects the outage after a few failed calls and skips the remote for the rest of the run, so the thousands of CAS loads and saves in a build instantly fall back to local compilation instead of each one paying a full timeout. Once the backend recovers, the cache resumes mid-build.
Behavior is unchanged for healthy caches: normal misses and successful uploads work as before, and a short-circuited save simply skips the upload, so the produced artifact is identical.