Indexes: ranking notes

No P1 in this category: the confirmed integrity bug (#35837) is narrow and self-signalling (it only misreports while the filter index is incomplete), and the framework decoupling work, while unblocking multiprocess, is durable-value rather than release-critical. Bands kept as originally assessed; only the ordering was rearranged to follow the category's stated priorities (integrity, then sync time, then decoupling, then features).

• Dependency chain: #34489 (batched index db writes) is stated groundwork for #26966 (parallel index sync); review #34489 first, and note #26966 already split its ThreadPool component out to #33689, so its remaining scope depends on both.
• #24230 and #35474 both move indexes away from node internals/global state for the multiprocess goal and touch overlapping call sites; #24230 is the larger, approach-setting change and should be decided first so #35474 does not lock in a conflicting ownership model.
• #35728 and #35837 are the same class of defect — index lookup failures flattened into 'not found' / 'completed' — so they set one error-propagation policy for index consumers; #35837 (common case: index still syncing) is the more frequent trigger and is ranked above #35728 (rare disk I/O errors) even though a single convention should cover both.
• #35827's realistic index benchmarks are the measurement tool for the claimed 11–27% and 6x gains in #34489 and #26966; it sits lower on its own merit, but a reviewer working on those two may want it in hand first.

This pass by claude-opus-5 on 2026-09-17 saw every PR in the category at once and checked the bands given to each PR alone against each other, ordered the PRs, and noted chains and overlaps. Back to the category.