[ScriptedAlchemy]

Finding

The canonical observation decoder issues one awaited SQLite query per distinct content hash, including from the existing batched session projection-observation read. A single observation with many large fields therefore serializes round trips to the same registered store, even though the body hashes are already known together.

Change

• Load canonical body rows in SQL IN batches of at most 16 distinct hashes. The existing observation-to-body authority, store identity, and transaction/snapshot still govern the read; this creates no second cache or worker.
• If an exact-SQL result exceeds its existing materialization ceiling, split the page, down to the original one-body operation. Other SQL failures remain typed errors.
• Decode and SHA-256/size-verify each body in the original sorted-hash order, independently of SQLite's result order. Missing and corrupt hashes still refuse. Repeated references share one lookup and remain hydrated through the canonical parser.
• No admission, persistence format, authorization, retention, mutation, or deadline policy changes.

Regression

An actual SQLite-backed TestConnection fixture builds one valid observation containing 18 distinct external bodies and a repeated body reference. A query-counting adapter on the real read port requires 2 bounded SQL reads instead of 18. It also checks the fully reconstituted observation and typed missing-body and hash-corruption refusals.

Verification / limits

• Independently reviewed the isolated one-file commit against current master adbf0e391bd7; GitHub comparison confirms one fast-forward commit changing only crates/tracedecay-session-temporal-store/src/query.rs. Checked error precedence, identity, deduplication, parameter bounds and materialization fallback at source level.
• Not executed: Bazel, rustfmt, Rust compilation, the added test, or runtime benchmarks. This environment has no supplied checkout or Bazel/Rust toolchain. Hosted CI needs to run bazel test //crates/tracedecay-session-temporal-store:unit_test --test_output=errors and formatting on this exact head. A query-count regression is not an elapsed-time speedup measurement.
• Separate PR/CI automation owns test repairs and merge decisions. Keep this draft until isolated-head validation passes.

This is a focused hot-path optimization found while reviewing open performance/storage issues; it does not claim to resolve the broader corpus, graph publication, or storage-growth incidents.

[changeset-bot]

⚠️ No Changeset found

Latest commit: acfc0b2

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

[ScriptedAlchemy]

Consolidated into #4742 (integrate/open-pull-requests) at the captured head shown in its description, as requested. Continue review and further fixes on #4742. This closure does not assert CI readiness; the replacement remains draft. Source branch retained.