start rate limit

The counter that ends a crash loop, set in [Unit]: five starts — StartLimitBurst=5 — inside the ten seconds of StartLimitIntervalSec=10s, and systemd stops bringing the service back. Start request repeated too quickly is then the last sentence on screen, so it is the one that gets blamed, and StartLimitBurst= gets raised. That is the consequence, not the cause. Active: failed (Result: exit-code) still names the real one, and the five actual errors — EADDRINUSE, 203/EXEC, a missing permission — sit above the counter lines in journalctl -u example-api -b: systemctl status prints only the last start attempt, and -n 100 will not bring them back, because the cut is a timestamp rather than a line count. Raising the limit buys you the same broken service crashing for longer and filling /var/log/journal, and since a unit stuck in automatic restart is not a failed unit, the outage leaves systemctl --failed. Fix the cause, then systemctl reset-failed.