[aft-alfonso]

Lands thirteen reviewed PRs as one merge, so they take one merge-queue round instead of thirteen. Each component PR stays open until this merges, then closes pointing here.

PR	Change
#1194	Views: import existing text and symbol indexes when views are turned on
#1200	Call graph: index inline-module candidates (cold build 636 s → 154 s on AFT, identical rows)
#1208	Call graph: prune workspace-member walks
#1201	Background commands keep their startup image across daemon exits
#1203	aft_inspect reports current diagnostics or partial, never stale (#829)
#1209	Lease heartbeats every 25 s instead of 5 s on idle projects
#1174	aft_callgraph resolves Python cross-file imports (#1126)
#1190	runon reattaches to an accepted job by id after a dropped connection
#1183	Killed and reclaimed commands skip output token counting
#1198	Flake batch 27 (test-only)
#1206	Bash startup deadline doesn't name a call as its own blocker
#1186	gh shim waits for Plexus's full reply window
#1191	gh shim reports Plexus pre-send refusals as not sent

Merge resolution:

• Views: import existing text and symbol indexes when views are turned on #1194 with Views: release a failed base-completion claim; call graph retries a busy owner #1154: view loads keep Views: release a failed base-completion claim; call graph retries a busy owner #1154's busy retry, and constructors use the import-aware builders. In aft_search, "import building" answers from the bounded scan, while a real load failure keeps the shared view_load_error_response.
• Views: import existing text and symbol indexes when views are turned on #1194 with Views: release a failed base-completion claim; call graph retries a busy owner #1154's completion_busy field: the import's unbound and import-building errors stay plain busy errors.
• runon: reattach to an accepted job by id after a dropped connection instead of reporting outcome unknown #1190 with runon: recovery allows the runner's real queue wait before declaring a job lost #1155: one outage fixture covers both started-or-queued jobs and the replay mode.

Gate on the build server (Linux):

• Formatting, the test-timing-hooks feature check and the Windows target check: clean.
• Integration binaries: 2,616 passed, 2 failed.
  • real_rust_analyzer_checks_do_not_block_cargo_build also fails on main there, because the build server sets CARGO_TARGET_DIR.
  • subc_bridge_health_check_returns_root_status_report passed 3/3 alone on this branch and 2/2 on main, so it was load.
• Lib: the secret-shell tests need the built aft binary, which that job didn't build. They're being rerun with it built.
• composition_no_marker_boundary_record: being run alone on this branch; it passes 3/3 on main.

The search descriptor is engine_unwired: the only fenced file is semantic_search/mod.rs, on the views-on import path.

---

^{Need help on this PR? Tag @codesmith-bot with what you need.}

---

Summary by cubic

Lands thirteen reviewed PRs as one merge so they take one merge-queue round instead of thirteen. Each component PR stays open until this merges, then closes pointing here.

Behavior changes

• Views: enabling views imports existing legacy text and symbol indexes; while an import builds, aft_search and grep answer from the bounded scan instead of failing (Views: import existing text and symbol indexes when views are turned on #1194).
• Call graph: cold builds index inline-module candidates per snapshot and prune workspace-member walks, cutting cold build 636 s → 154 s on AFT with identical rows (Call graph: index inline-module candidates instead of scanning every Rust file per reference #1200, Call graph: prune workspace-member walks and share canonical probes per build #1208).
• aft_callgraph resolves Python cross-file imports including aliased and relative forms, and discloses matching unresolved callers in nonempty answers (aft_callgraph: resolve Python cross-file and function-local imports (#1126) #1174, aft_callgraph never resolves cross-file Python call edges; callers under-report and disclosure only covers zero-caller answers #1130).
• Background commands keep a hard-linked copy of the daemon's startup image, so a daemon exit can't break a sandbox still launching (Keep background commands' startup image alive across daemon exits #1201).
• aft_inspect reports current diagnostics or the file as partial, never stale results after a change (aft_inspect: report current diagnostics or partial, never stale (#829) #1203).
• Idle project leases heartbeat every 25 s instead of 5 s (Lease heartbeats: rewrite every 25 s instead of 5 s on idle projects #1209).
• runon reattaches to an accepted job by id after a dropped connection (runon: reattach to an accepted job by id after a dropped connection instead of reporting outcome unknown #1190).
• Killed and reclaimed commands skip output token counting (Shell: killed and reclaimed commands skip output token counting #1183).
• Bash startup deadline no longer names a call as its own blocker (bash startup deadline: don't name a call as its own blocker; rebind wait diagnostics #1206).
• The gh shim waits for Plexus's full reply window; pre-send refusals report as not sent, and only transport_unknown and lease_expired get one same-nonce retry (gh shim: wait for Plexus's full reply window before reporting a bot write's outcome unknown #1186, gh shim: adopt Plexus's not-sent pre-read refusals (fixtures at fec01fef) #1191).

Merge resolution and tests

• Views: import existing text and symbol indexes when views are turned on #1194 keeps Views: release a failed base-completion claim; call graph retries a busy owner #1154's busy retry and completion_busy handling; runon: reattach to an accepted job by id after a dropped connection instead of reporting outcome unknown #1190 shares one outage fixture with runon: recovery allows the runner's real queue wait before declaring a job lost #1155.
• Flake batch 27 lands test-only fixes: dirty-page VFS method copies, inspect dedup deadline, and storage-scoped v3 free-space test overrides (Flake batch 27: dirty-pages VFS copies, inspect dedup deadline, storage-scoped free-space overrides #1198).

^{Written for commit 4dd724f. Summary will update on new commits.}