fix(entities): run remaining timer callbacks on the event loop

[buggedcom]

Problem

Recent Home Assistant cores promote non-thread-safe state writes from a warning to a hard RuntimeError, which crashes the update. Three timer handlers were registered with async_track_time_interval / async_call_later as plain functions, so HA runs them in a worker thread:

• binary_sensor.py — _on_time_interval → calls async_write_ha_state() (this is the crash users are hitting in the field)
• sensor.py — _handle_scan_tick → calls async_create_task() (loop-only)
• sensor.py — _start (the async_call_later callback) → calls async_track_time_interval / async_on_remove (loop-only)

Observed error:

RuntimeError: Detected that custom integration 'hass_datapoints' calls
async_write_ha_state from a thread other than the event loop ...
binary_sensor.py, line 127

Fix

Mark all three handlers @callback so HA schedules them on the event loop, where these calls are safe. All three do only non-blocking, loop-affine work, so @callback is correct (no blocking I/O moved onto the loop).

The store-listener path (_on_store_update) was audited and is fine — _notify_listeners() is only ever invoked from loop coroutines.

Relationship to #11

#11 fixed the same bug class for sensor._handle_time_interval. This completes the sweep for the remaining three async_track_time_interval / async_call_later registrations.

Also

Bumps manifest.json version 0.5.1 → 0.6.1 (it had drifted behind the released tags).

🤖 Generated with Claude Code