Badge with error count is incorrect

[elliette]

Re-visiting a screen that has a RenderOverflow causes the error badge count to increase, which is confusing since when you navigate to the inspector screen you see only one error.

Furthermore, visiting the inspector screen causes the error badge to disappear, even though the error still exists.

[khanak0509]

Hey @elliette! I was working on a fix for this and ran into kind of a weird edge case. _handleConnectionStart() runs the first time we open the tab and basically wipes all the errors.
because of that, if an error happens while we are on something like the Performance tab, then the first time we switch to the Inspector it just disappears, even though it’s still valid. After that it behaves normally. now I just wanted to check, should we avoid clearing errors in _handleConnectionStart() so they persist on first load, or is there a specific reason we want to reset everything when the Inspector initializes?

[ibrahim-iqbal]

Looks like this was already fixed by #9805 (merged June 24) — the PR description says fixes issue #9524 and the diff directly addresses both symptoms described here:

• badge disappearing on revisit: clearErrorCount/clearErrors now skip screens with _activeErrors tracked (so the Inspector's managed count survives tab switches), and InspectorController calls clearErrors once on connect instead of on every visit.
• count increasing incorrectly: incrementBadgeCount now returns early for screens with an active-errors map (avoiding the blind double-increment), and appendError only increments/decrements based on the error's actual read-state transition rather than unconditionally.

Checked current main and the fix is present. Looks safe to close unless there's a follow-up regression — happy to be told otherwise.