diff --git a/quest/m0/nightly/relay-join-regressions.md b/quest/m0/nightly/relay-join-regressions.md deleted file mode 100644 index e7fba628d1..0000000000 --- a/quest/m0/nightly/relay-join-regressions.md +++ /dev/null @@ -1,30 +0,0 @@ -# [M] The 2026-11-20 relay session regressions are fixed or accepted - -## Goal - -Nightly's Relay session benchmarks pass the paired comparison again: each -regression from 2026-11-20 is fixed, or accepted by the PR that caused it -with a stated reason. - -## Plan - -Found in run [44767045952](https://github.com/moq-dev/moq/actions/runs/44767045952), -`a759e4cf6` against `a4fa80a74`. Six regressions, consistent across all -three rounds, all on `moq-lite-07-wip`: - -- `relay_session_join_broadcasts/*` +12 to 20%, at 1 to 8 relays and 64 or - 1024 broadcasts. -- `relay_session_join_relays/relays=8/publishers=16` +14.8%. -- `relay_session_delivery_paced/relays=1/publishers=16/viewers=256` +26.9%. - -Suspects, the lite-07 route and subscription changes in that range: -[moq-dev/moq#9112](https://github.com/moq-dev/moq/pull/9112), -[moq-dev/moq#9155](https://github.com/moq-dev/moq/pull/9155), -[moq-dev/moq#9174](https://github.com/moq-dev/moq/pull/9174), -[moq-dev/moq#9179](https://github.com/moq-dev/moq/pull/9179), -[moq-dev/moq#9183](https://github.com/moq-dev/moq/pull/9183), -[moq-dev/moq#9202](https://github.com/moq-dev/moq/pull/9202), and -[moq-dev/moq#9219](https://github.com/moq-dev/moq/pull/9219). Bisect with a -paired local run of one `join_broadcasts` cell before changing code. - -Public API: none expected. Wire: none.