Task improvements

[bjorkert]

Improvements

• Throttled Scheduling:
  Background refresh notifications are now re-scheduled at most once every 10 seconds, preventing intensive calls to this functionality.

• Device Status Retry:
  When device status returns empty, the system now retries after 5 minutes instead of giving up immediately.

• Alarm Check Delay:
  Alarm check tasks now wait 5 seconds after .deviceStatus, .treatments, and .fetchBG are complete before re-attempting, avoiding immediate re-triggers.

[marionbarker]

Code Review

Looks good

Test

Built onto test phone and confirm proper display for URL fed from Loop, Trio 0.2.3 and Trio-dev

• success: Loop
• success: Trio 0.2.3
• success: Trio-dev (private beta)

Observation, the prediction lines take a cycle to show up for both Loop and Trio 0.2.3, but appear immediately for Trio-dev. Not sure if that was always true.

Built onto personal phone as well.

[marionbarker]

Approved based on code review and testing.

[marionbarker]

Test with "problem" URL

Get the URL that stimulated this change (person doesn't have device status data)

• Build PR 379 version to Mac.
  • Open the URL that was giving the user a problem
  • Confirm it opens properly
• Build v2.2.10 to Mac and observe high rates of logging:

QUARANTINED DUE TO HIGH LOGGING VOLUME
Type: Fault | Timestamp: 2025-03-06 09:49:19.121711-08:00 | Library: libsystem_trace.dylib | Subsystem: libsystem_trace.dylib | Category: Default | TID: 0x673e951

• Build v2.2.10 to a real phone and observe LoopFollow crashes for this URL
• Build PR 379 version to real phone and now LoopFollow opens properly

Observation - restored real phone to my personal Loop URL and notice that the forecast displays promptly this time.