perf(server): sidebar shell loads 3x faster and startup thread counts 80x faster - #27106

Eligibility
This optimizes sidebar snapshots and startup reads by fetching only relevant thread shells and counting threads directly in SQL. Existing subscriber options and operation scopes remain unchanged, with reported snapshot-equivalence checks supporting the limited behavioral risk.

Additional Thinking

Your Custom Policy

Applied default criteria. The assessed changes neither change product defaults nor add or broaden static-analysis suppression directives, so neither additional human-review condition is triggered.

Code

Replaces full-shell reads with a lightweight active-thread index, targeted shell queries, and a SQL thread count. HTTP/WebSocket snapshots retain subscriber opt-ins and read consistency through a read-only transaction. Project deletion and branch rename move existing filters into queries rather than broadening their targets; relay startup avoids subagent shells.

Review Context

Author is a repository admin and ping-devs member; no ownership requirement was supplied. No unresolved review comments, correctness findings, blocking-severity configuration, or check results were supplied. The PR reports byte-identical snapshots across all four opt-in combinations on a static database and measured improvements on 2,923 threads; these support the limited performance-refactor intent. Seven ignored files were left outside the assessment.

Verdict

The understood change reduces read and decoding work without introducing new persistence operations, publication permissions, or product choices. Existing destructive callers retain their selection scope and guards. Transactional snapshot reads and reported output-equivalence evidence support approval without a material decision requiring human judgment.

Notes:
No code owners found in target merge base 08a463e

Agent Credits: 30 credits
Approvability runs at a flat rate of 30 credits whenever the agent is asked for a verdict.