You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Known limits left by the issue batch (Claude follow-up answers, config snapshot size, large trees, unmeasured memory) #395
Known limits recorded by pull request 387. Each needs a design decision or evidence that the pull request did not have.
Claude's own answer after a background agent finishes during a turn is not tracked as work. Only the idle timeout protects it, and a revert is not refused while it runs a tool. Three designs to track it were written and removed after review found defects in each (a count settled by the wrong event, a count that sticks, a revert allowed under a running tool). Claude's result names the kind of event but not the task it answers, and a count cannot tell waiting for output from running a tool. docs/providers/claude.md lists the cases with the captured Claude Code 2.1.326 record sequences.
The config stream sends a full snapshot for a capability change or a one-provider change. With a real Codex account a snapshot is about 1 MB. Sending less needs a new wire event.
File trees past the 25,000-entry index cap stay partial. That needs per-directory loading.
Issue 390's memory criterion was not measured. Replies are bounded per call; that desktop memory returns to baseline, and that it was retained at all, are unproven: the test repositories are small.
Issue 381's root cause is unknown. What makes WebKit leave an indexedDB.open unanswered is not established; the app now shows an actionable error instead of loading forever. It was seen once more in a packaged run on the final build.
The packaged UI suite cannot resize its window under a tiling Wayland compositor, also with a floating rule; it runs under Xvfb there.
A commit with tens of thousands of parents can exceed the history page cap; cutting parents would break the graph.
Closing as not planned: each item needs a new design decision or evidence that is not available (a new wire event, per-directory loading, a memory measurement on a large repository, the WebKit cause). Recorded here so each can be reopened as its own issue with that decision or evidence.
Activity
mubeda commentedon Nov 19, 2026
Closing as not planned: each item needs a new design decision or evidence that is not available (a new wire event, per-directory loading, a memory measurement on a large repository, the WebKit cause). Recorded here so each can be reopened as its own issue with that decision or evidence.