Harden the heartbeat thread

Keep the beat loop alive across unexpected errors, and round the derived
stale timeout up so a sub-second interval can't produce 0.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

Author: snopoke

taskbadger/_heartbeat.py

@@ -81,7 +81,14 @@ def _ensure_thread(self) -> None:
 
     def _run(self) -> None:
         while True:
-            self._wake.wait(self._beat())
+            try:
+                timeout = self._beat()
+            except Exception:
+                # Never let the thread die: every registered task would then go
+                # stale with nothing to restart the pings.
+                log.exception("heartbeat beat failed")
+                timeout = 1.0
+            self._wake.wait(timeout)
             self._wake.clear()
 
     def _beat(self) -> float | None:

taskbadger/_integrations.py

@@ -12,6 +12,7 @@
 
 import collections
 import logging
+import math
 import re
 
 from . import sdk
@@ -131,7 +132,9 @@ def resolve_heartbeat_options(heartbeat_interval, stale_timeout, system):
         heartbeat_interval = None
 
     if stale_timeout is None and heartbeat_interval:
-        stale_timeout = int(heartbeat_interval * STALE_TIMEOUT_FACTOR)
+        # `stale_timeout` is whole seconds, so round up: a sub-second interval
+        # must not produce a timeout of 0.
+        stale_timeout = max(1, math.ceil(heartbeat_interval * STALE_TIMEOUT_FACTOR))
 
     return heartbeat_interval, stale_timeout