From 7c0f589bbe4561bec0cbc0bbb3248a2a55a6d6c5 Mon Sep 17 00:00:00 2001 From: Phil Windle Date: Mon, 15 Jun 2026 15:53:07 +0000 Subject: [PATCH] test(ci): mark e2e_epochs/epochs_mbps_redistribution as flaky The test asserts that all late txs land in the checkpoint's last block, but the proposer seals the last block as soon as one tx is available and snapshots the mempool, so late txs still propagating over gossip spill into the next checkpoint. This is a timing race in the sequencer, not a redistribution bug; a follow-up makes the last block collect txs until its build deadline. Until then, flag the test as a flake so it does not block the merge queue. --- .test_patterns.yml | 7 +++++++ 1 file changed, 7 insertions(+) diff --git a/.test_patterns.yml b/.test_patterns.yml index 498fc75e22ad..3816c919c8e4 100644 --- a/.test_patterns.yml +++ b/.test_patterns.yml @@ -275,6 +275,13 @@ tests: owners: - *sean + # Asserts all late txs land in the checkpoint's last block, but the proposer seals the last block as + # soon as one tx is available and snapshots the mempool, so late txs still propagating over gossip + # spill into the next checkpoint. Flaky until the sequencer waits for the last block's tx window. + - regex: "src/e2e_epochs/epochs_mbps_redistribution.test.ts" + owners: + - *phil + # Blanket flake patterns for unstable p2p, epoch, and l1 tx utils tests # This is a temporary measure while team bandwidth is constrained # Replaced many specific patterns - see https://github.com/AztecProtocol/aztec-packages/pull/17962 for historical context