[thebhef]

The plot asked for bus-error episodes only from an area's resample and
from a gap-setting change. The resample loop runs only while
live-following, so a plot that was stopped, paused, or scrubbed back
into history never asked again, and new errors appeared only after a
pan or zoom (owner, on PEAK hardware). Even when something did ask,
useBusErrorMarkers memoised a request for an unchanged window after a
complete answer, so errors landing inside that window were dropped.

Both the trigger and the freshness token are now a model fact rather
than a timer: the host's per-bus errorCount from useBusHealth(),
which re-reads on bus-health-changed. The host's emitter publishes on
its poll tick whenever any health row changed, error count included,
so a count move always reaches the view.

• PlotPanel sums the error count over the project's buses and, when
  it moves, asks again for the unmoved window — the same effect shape
  as the gap-setting refetch beside it.
• BusErrorMarkerRequest gains errorsSeen: part of the memo key, not
  of the invoke arguments, so a fault since the last answer makes
  the same window a new request.
• A complete: false answer was already never memoised; unchanged.

Tests: a hook test that the same window with a higher errorsSeen is
asked again (and the count is not sent), and a PlotPanel test that a
bus-health-changed with a higher count, after mount has settled,
issues exactly one new bus_error_episodes_in_window with the same
from/to and no area resample. Removing either half of the fix turns
them red. README's error-frames passage says markers refresh as errors
arrive.

Co-Authored-By: Claude Opus 5.5 noreply@anthropic.com

[thebhef]

Merge activity

• Oct 14, 9:10 PM UTC: A user started a stack merge that includes this pull request via Graphite.
• Oct 14, 11:29 PM UTC: Graphite rebased this pull request as part of a merge.
• Oct 14, 11:32 PM UTC: @thebhef merged this pull request with Graphite.