[MikeBishop]

Sporadic "No internet connection." messages could blank the calendar after a single transient network error, and the message could linger after the connection came back.

• Track fetch errors per calendar URL instead of one error per module, so one failing calendar no longer hides the others once they have recovered.
• Clear the error and redraw as soon as a calendar fetches successfully again, even if the events checksum is unchanged. Previously the redraw was skipped, so the error stayed until the next once-a-minute self update.
• Include url in the CALENDAR_ERROR socket notification so the front end can attribute errors to a calendar.
• Add an opt-in keepEventsOnError option (default false, so current behavior is unchanged). When true, the last known events stay visible after a failed fetch, and the error is only shown if there are no events to display.
• Add unit tests for recovery with an unchanged checksum, multiple failing calendars, the default behavior, and keepEventsOnError.

No related issue.

🤖 Generated with Claude Code

[KristjanESPERANTO]

Thanks for the PR!

Speaks something against enabling this behavior permanently and omit the new opt-in option?

[MikeBishop]

Considered that, but it means that the error isn't visible from the screen unless all calendars have failed from the start. Some people might prefer to see an error so they can fix it. Some visual cue that the shown events are stale might be an alternative.

[KristjanESPERANTO]

Good point. What do you think of using a small grace period instead of a new config option - e.g. keeping existing events for the first 2 or 3 consecutive fetch failures and only showing the error if it persists? That would handle transient glitches by default while still surfacing permanent errors.