[SandObserver]

Summary

• Fetch the container list for GET /api/health through latest(): one request to the socket proxy at a time, a wait of at most BATCH_MS, then the last list up to 2 minutes old.
• Start each item's ping before waiting for the container list, so both run at once.

Testing
All green. Added 1 test: a socket proxy that never answers delays the health reply by at most the batch timeout and still reports the ping.

Changelog

Fixed

• Keep app health dots updating while the Docker socket proxy stops answering.

Checklist

• Stays within the project constraints: one container, no runtime dependencies, no build step (see CONTRIBUTING.md).
• Tests added or updated for new logic, and the full suite, lint, and typechecks pass.
• Docs updated if behavior or configuration changed.