A paused service is now left out of the "servers are down" alerts.
Pausing a service is usually the prelude to taking its backend down and working on it — at which point its servers fail their health checks, and the watchdog would email that they were down, telling you what you just did. A service in maintenance mode is now skipped by those health-check alerts entirely. When it is resumed and its servers pass again, the next round reports from wherever it left off, so a genuine problem after resume still alerts and a clean resume sends no false "recovered" message. (The URL probe already treated a paused service's clean 503 as an answer, so it never contributed to this — only the health-check alerts did.)
Upgrading from 1.91.0: nothing to do.
Packages are attached below; the container image is at ghcr.io/avandeputte/haproxy-manager:1.91.1.