6.6 Build Service Operations

Use this when a build crosses CI, Bazel, cache, remote execution, BES, and downstream consumers and no single log explains the failure or latency.

Escalate from portable Bazel and REAPI evidence to governed private executor signals inside an opaque action without contaminating cached results.

Use this when a single average build time or service CPU graph cannot tell whether users are waiting on lookup, transfer, queueing, execution, outputs, or missing results.

Use this when “the remote build failed” could mean source, Bazel, credentials, cache storage, scheduling, execution, network, BES delivery, or a downstream consumer.

Use this when deciding whether a failure is deterministic, flaky, incompatible, transient infrastructure, or corrupt shared state—and whether retry will help or amplify it.

Use this when cache, remote execution, BES, region, or executor-fleet incidents require recovery without destroying evidence or causing a cold storm.

Use this when sizing cache, scheduler, executor, BES, and CI stages for baseline, bursts, scale lag, and component failure.

Use this when deciding whether cache, remote execution, warm CI, or local investment delivers real value instead of moving cost elsewhere.

Use this when measurement identifies loading, analysis, configuration, retained-state, hashing, transfer, materialization, or execution as the bottleneck and you need the matching bounded mechanism.