perf: stop hydrating score telemetry in the allocator floor read

[Peyton-Spencer]

Summary

Backport of ditto-assistant/ditto-subnet#559. Production is served from this repository, so the change has to live here to take effect.

Based on fix/emission-catchup-between-waves (#673) rather than main, because prod is currently running that branch — this keeps the deployed commit moving forward instead of reverting it. Review the top commit only; #673 covers the one below it.

get_score_priority_floor_rows passed include_fingerprints=False but not include_details=False, so every eligible ledger row arrived carrying its full per-case audit blob (~22KB of JSONB across ~1.2k rows on bench v8) which the allocator immediately discarded.

Why now — live production evidence

• ditto-api pinned at 100% CPU on its single worker
• POST /validator/job: 7,476 × 204 against 24 × 200 in one log window — 99.7% of validator polls return no work
• slowest observed /validator/job: 40,181ms (subnet#388 filed p95 at 11.6s)
• the continual-retest lane is not gated — 34 × 200 issued against 2 × 409. It is starved of turns on the same saturated event loop, which is what the uneven shared-seed counts on the leaderboard actually reflect.

Safety

Nothing downstream reads the field, verified before changing it:

• rank_submissions and resolve_ranking_scores never touch telemetry
• get_score_priority_floors returns bare floats
• the only caller that keeps a floor row (_public_submission_pipeline) reads row.agent_id
• no .row.details access exists anywhere in the tree

list_eligible_ledger already documents include_details=False as correct for "queue-floor, cleanup, and efficiency-cohort consumers" — this consumer was simply missed.

Validation

• uv run pytest ditto/tests/db/queries/ -q → 628 passed, 4 skipped
• uv run pytest ditto/tests/api_server/endpoints/test_validator.py -q → 220 passed
• ruff check / ruff format --check / mypy → clean

The new test seeds rows that do carry details, so details is None distinguishes never selected from nothing to select.

Scope

This does not close subnet#388. The batched idle-slot claim contract (POST /validator/jobs), the shared per-batch AllocatorPrioritySnapshot, and the validator-side rolling-upgrade negotiation are all still unimplemented — that is the change that removes the per-slot recomputation, and it spans both this repo and the validator in the monorepo.

[Peyton-Spencer]

Closing unmerged: superseded by the monorepo cutover.

Production is now served from ditto-assistant/ditto-subnet (/opt/ditto-subnet/apps/platform), so this repository no longer deploys anywhere. This PR's content already landed in the monorepo and is live:

• fix: keep emission catch-up running between waves #673 → ditto-subnet#554 (emission catch-up between waves)
• perf: stop hydrating score telemetry in the allocator floor read #674 → ditto-subnet#559 (allocator floor read stops hydrating score telemetry)
• perf: project only the KOTH detail keys the retest path reads #675 → ditto-subnet#563 (KOTH detail-key projection — took /validator/job from 20-40s to 438-1247ms)

Merging here would land the fix on a branch nothing ships, which is worse than closing: it makes this repo look maintained and invites a deploy from it. See ditto-subnet#566 for the cutover follow-ups.