[ScriptedAlchemy]

Superseded by #4742. Commit e28de398 includes this PR's complete head b484bc84 as its second parent and retains both source files byte-exact. The source branch remains intact. This closes duplicate review scope only; the imported draft source is not claimed validated, fixed for #4558, or merged into master. #4558 remains open.

Root cause

Refs #4558. The earlier merged #4709 reserves response-settlement time for project graph reads, but a separate callable graph path still spends almost all of its remaining request budget waiting for a released engine during seed classification and then again for the catalog. The first tracedecay_callers/graph navigation after a release can consequently reach the enclosing MCP dispatch timeout before returning an unavailable/retryable state.

Change

• Reuse the existing measured rewarm_wait_budget policy from project graph reads; expose it only within the crate.
• Probe readiness with a zero-budget check, which starts background re-warming if needed. If measured remaining warm time does not fit with the reserved half-budget, return the callable's existing typed Unavailable result promptly instead of blocking through the dispatch deadline.
• Apply the bounded path to both seed point classification (engine) and catalog readiness; retain the engine lease during the actual point lookup.
• Reject a request that no longer has an admitted generation-resolution budget rather than attempting an unretained reader open.

Regression coverage

Adds slow_callable_rewarm_reserves_time_for_seed_and_catalog_reads to exercise the shared budget with a 10-second outer deadline, a measured 17-second cold warm, and a short 2-second warm. The existing a_callers_request_after_park_waits_for_the_graph_to_rewarm integration journey covers the short/ready case, but was not run on this branch.

Validation and limitations

The exact two-file commit diff was independently read against master 02e6467. Rust/Bazel build, formatting and tests could not be executed in this tool environment; there is no claimed test pass or verified 6,219-file timing improvement. Hosted current-head CI, compile and actual release-to-first-call integration are required. No authorization, deadline, integrity, or admission limits are increased.

This is partial progress on #4558. The callable API currently maps the rejected wait to typed Unavailable, not a dedicated Rewarming { retry_after_millis } response. The original 6,219-file release/10-second MCP reproduction and, if needed, precise retry-after propagation remain outstanding. Keep #4558 open.

The larger unrelated integration PR #4742 and release PR #4744 are untouched; the separate PR-repair automation owns CI, repairs, and merge.

[changeset-bot]

⚠️ No Changeset found

Latest commit: b484bc8

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