Verdict: CI infrastructure failure, not a test or PR bug.

The Unit Tests / Rust Unit Tests (gke-runners-amd64, x86_64, x86_64-unknown-linux-gnu) job never got to run any test. It was still downloading build artifacts (it had just finished feldera-test-binaries-x86_64-unknown-linux-gnu and was about to start feldera-sql-compiler) when the runner itself was shut down:

##[error]The runner has received a shutdown signal. This can happen when the runner service is stopped, or a manually started runner is canceled.
Cleaning up orphan processes

• Job: https://github.com/feldera/feldera/actions/runs/41181351431/job/124483863469 (log line 379)
• This was the only job GitHub reported as failure; every other job in the run shows cancelled, i.e. collateral from this one job dropping out.
• The run was a Merge Queue Tasks run for PR [adapters] Keep the JSON text of floats stable #7735 (fred/observe-7735-41181351431 investigation branch), which only touches crates/feldera-types, crates/sqllib, crates/static-connectors — nothing related to CI runners or artifact download.

Next step: likely a transient GKE runner eviction/node issue. Re-queueing the PR should pass. Flagging so a pattern can be tracked if gke-runners-amd64 keeps getting shut down mid-job.