[ScriptedAlchemy]

Root cause

Refs #4532. The current sealed graph reader already pins the last segment, avoiding a new open() for consecutive pages from the same file. However, it drops that handle when a different segment is encountered. An interleaved A, B, A, B page sequence therefore reopens a segment on every switch, which is expensive on the reported endpoint-scanned macOS machines.

Focused change

• Extend that existing, pass-local pinned handle into an 8-entry least-recently-used handle window. Revisited segments retain the same opened identity; the oldest handle closes on eviction, and an evicted segment must pass the original open-time checks again.
• Continue using the original exact range validation, opened-file size checks, full graph-page digest verification, and the per-read cancellation/deadline probe. The change does not retain handles between passes or introduce a process-wide cache.
• No admission, store format, durable ownership, auth, deadline, or resource budget changes.

Regression coverage

• A real two-segment filesystem fixture invokes the production with_verified_segments path, alternates six page reads, verifies returned content/digests, and requires 2 opens rather than 6.
• A separate filesystem regression cycles through more than 8 distinct segment files, verifies the handle bound, refuses an out-of-bounds read on a cached segment, and requires a fresh validated open after eviction.
• Existing sealed graph tests still cover mutation/unlink, corrupted payload, cancellation/deadline precedence and the ordinary contiguous range-reader path.

Validation

Reviewed the exact one-file GitHub diff against master at 65ddc060 (133 added, 11 removed) and checked source structure, whitespace and final newline. Rust compilation, rustfmt and the two new tests have not run: this execution environment has no supplied checkout or Bazel/Rust toolchain. Hosted CI must run the isolated branch through bazel test //crates/tracedecay-store-runtime:unit_test --test_output=errors and the graph publication journey before merge. No measured wall-time or CPU improvement is claimed.

Remaining acceptance

This is a focused remaining reopen case, not a substitute for the issue's required before/after graph-publication timing and file-open measurements on the reported 6.6k-file macOS endpoint-security setup. The existing contiguous page reader already opens one handle per segment. Keep #4532 open until the actual production journey is verified. This PR is independent of the graph rewarm PR #4709; PR repair/CI/merges remain owned by the separate PR automation.

[changeset-bot]

⚠️ No Changeset found

Latest commit: e06fc29

Merging this PR will not cause a version bump for any packages. If these changes should not result in a new version, you're good to go. If these changes should result in a version bump, you need to add a changeset.

Click here to learn what changesets are, and how to add one.

Click here if you're a maintainer who wants to add a changeset to this PR