[LeeroyHannigan]

What

Four design documents and the technical debt register described a PostgreSQL read replica mode as current behaviour: a storage.postgres.read_replica_url setting, a read_replica_pool_size, a read_pool() selected per read, and ConsistentRead=false reads routed to a streaming replica. None of that is in the repository. The PostgreSQL backend serves every item read and write from data_pool, and PostgresStorageConfig carries #[serde(deny_unknown_fields)], so a configuration that copies read_replica_url from the documents stops the server at startup.

The documents now state what the backend does and label the replica mode a proposal.

• docs/design/01-requirements.md: REQ-DATA-015 states that the parameter is accepted and charged and that no in-tree backend routes a read on it. REQ-PG-008 is marked proposed, names the struct that refuses the keys, says every item read and write runs on data_pool, and points at both proposals (document 04 section 5.3 with one replica URL, document 11 Stage 2 with a replicas list).
• docs/design/02-high-level-design.md: section 4.7 states the decision as the PostgreSQL backend serving every item read from the pool that takes the writes, with the replica routing as a proposal; the rationale bullets are rewritten in that tense.
• docs/design/04-component-storage.md: the section 5.2 sketch carries a label saying it is a proposal, that its pool predates the catalog and data pool split and stands for today's data_pool, and that the two keys stop the server. Section 5.3 is left to feat(storage): carry ConsistentRead to the data engine as ReadOptions #559, which rewrites it.
• docs/design/08-component-config.md: the read_replica_url environment variable row and the two sample lines are removed, since a configuration carrying them does not load.
• docs/technical-debt.md: the F-2 row says the two ConsistentRead values are served the same way on every backend, and its line reference is corrected to item.rs:176.

Why

Closes #561.

An operator who followed document 08 got a server that refused to start. A backend author reading documents 01, 02 or 04 would have built against a read_pool and a configuration field that do not exist. The TODO at crates/core/src/types/item.rs:176 is the only trace of the mode in code and it is future tense.

Testing done

Documentation only; no code or test changes. Each statement in the new text was checked against the source at 744ea291: git grep read_replica and git grep read_pool under crates/ return nothing; deny_unknown_fields is on PostgresStorageConfig in crates/storage-postgres/src/config.rs; get_item_impl, query_impl, scan_impl and the item writes use self.data_pool; the capacity charge is in crates/engine/src/capacity_helpers.rs. The new text does not depend on #559: it names neither ReadOptions nor ADR-0024, and git merge-tree of this branch with feat/read-options produces a clean tree in either order.

git grep -n -E 'read_replica|read_pool' -- crates/   # no matches
git diff 744ea291 HEAD --stat                        # 5 files, 15 insertions, 15 deletions

Checklist

• I have read CONTRIBUTING.md
• All tests pass (no code change)
• Code is formatted (no code change)
• Clippy is clean (no code change)
• I have added or updated tests for new functionality (none needed)
• I have updated documentation if behavior changed
• Breaking changes are noted below (none)
• If this changes the wire protocol, Storage trait, auth model, on-disk
  format, or public CLI surface, an RFC has been accepted or is linked
  below. Otherwise, an ADR captures the decision (link below).

ADR / RFC: n/a (documentation correction; the decision the documents now describe is the one in the code)

---

By submitting this pull request, I confirm that my contribution is made under
the terms of the Apache License 2.0 and I agree to the Developer Certificate of
Origin (DCO). See CONTRIBUTING.md for details.