Warning: This build has drifted.
The coverage report for this pull request build may be inaccurate because its base commit is no longer the HEAD of its target branch.
This means it includes changes from outside the original pull request, including, potentially, unrelated coverage changes.

    • Learn more: For more information on this, see Tracking coverage changes for pull request builds.
    • Fix now: For a quick fix, rebase this PR at GitHub. Your next report should be accurate.
    • Prevent going forward: To avoid this issue with future PRs, see these Recommended CI Configurations.

C3 residual — the real outstanding bug. Two separate quadratic paths, both in the ETS/Mnesia claim/read hot path:
1. TableEngine.events_from_sequence/5 and read_partition/5 used a guarded select/2 then Enum.sort_by then Enum.take(limit). On an ordered_set that traverses and materialises every later event to keep limit of them. Replaced with an ordered key walk from a probe key, which needed a new next_key/2 on the Table.Store behaviour (:ets.next / :mnesia.dirty_next).
2. claim_one/7 restarted the scan from the head of open_deliveries on every lease, so claiming N leases re-walked the growing already-leased prefix. Replaced with claim_batch/8, a single ordered pass that carries its position across claims. This also deleted the duplicated claim_many/claim_loop/claim_one trio from both ets.ex and mnesia.ex (−38 lines each), continuing the M1 de-duplication.
Measured, same mix rheo.bench:
events	v0.11.0	v0.11.1 committed	now
1,000	655/s	21,980/s	70,681/s
6,000	—	11,846/s	72,863/s
24,000	—	—	60, to 24k instead of decaying ~4× between 1k and 12k.
fetch(limit: N), before/after on the same machine (µs per lease):
limit	committed HEAD	now
100	63.6	14.3
1,000	539.1	9.7
2,000	1,119.2	10.6
4,000	5 s timeout → {:error, :backend_unavailable}	~11
That last row was a latent production bug, not just slowness: a single large fetch exceeded the 5 s ets_call_timeout. The documented max_demand: 100 example was already paying 6× the necessary cost.
Also fixed
- L12 — Rheo.Producer.drain/2 now returns {:error, :already_draining} instead of silently overwriting a pending drain's caller, matching Rheo.Group.drain/2. Spec and doc updated.
- M8 — Rheo.LiveDashboard.Page uses the stateful row_fetcher form ({&fetch_health/3, nil}) with a 2 s TTL cache, so re-sorting or re-paging no longer replays the list_streams + N×list_groups + M×group_info walk.
- H5 — new test/rheo/telemetry_metrics_test.exs and test/mix/tasks/rheo_bench_test.exs. No file is at 0% coverage any more. The Met... (continued)