Release notes
Zero-config health observability
Services deployed on Sylphx can expose a continuous health score instead of a binary up/down probe. The score combines event-loop lag, queue depth, error rate, memory pressure and dependency pings, and it drives drain/restart decisions at the runtime boundary with no app-side wiring.
What changed
What this entry shipped
- Middleware for the common server frameworks, mounted like any other handler
- Continuous score instead of a binary check, so a draining service can be told apart from a dead one
- OpenTelemetry-native metrics, so the signal lands in the tooling you already run
- Audit chain output for compliance evidence
- Documented in the docs under Health
This entry publishes at /changelog/2026-05-20-zero-config-health-observability. The whole record is at /changelog, and the same entries are available as a feed at /changelog/feed.xml.
Correction: Entry wording tightened, figures withdrawn. The text of this entry is one of the changes the September 2026 correction records — read the correction.