[koabula]

Why

CI on main failed after #3 in leakage_corpus::only_a_payloads_size_reaches_the_host, though the same code passed on the PR (run 42968815830):

b_accept/Deferred("publish_shared_state_request")/body#json.devices[].keypackageRef.createdAt carries {Identity(Alice), Identity(Bob), Body(Alice), Body(Bob), AttachmentName, AttachmentLen}

The corpus ran its three replicate runs first and back to back. A KeyPackage's createdAt, notBefore and expiresAt are whole seconds (src/mls_adapter/mod.rs, now_seconds * 1000). When the three replicates fell inside one second and a later run did not, those fields held still across the replicates and differed elsewhere, so the classifier read them as carrying every secret. Whether the test passed depended on where the second boundaries fell.

What

matrix() now places the replicates at the first run, the last and one in the middle, and returns their indices. Runs execute in order, so a value read off a coarse clock either differs between the two ends and is fresh, or is the same in every run. Constant verdicts cannot change, because they depend only on the values and not on which runs are replicates; only a clock-dependent field can move from carrying to fresh.

Test code only: the module is #[cfg(test)].

Not verified locally

No Rust toolchain on this machine; CI is the first build.

🤖 Generated with Claude Code