mailrs-delivery-executor 1.2.0

Group-commit delivery executor on top of mailrs-maildir 1.2's deliver_batch. Accumulates per-path delivery requests from concurrent async tasks (SMTP / LMTP / IMAP APPEND sessions) and flushes each path's batch via a single fsync, with optional pipelined in-flight flushes (max_concurrent_flushes=2 default) to hide disk wait behind batch collection. Caller awaits a oneshot::Receiver for per-message confirmation. Default tuning: max_batch=64, max_wait=10ms, max_concurrent_flushes=2.
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23

# mailrs-delivery-executor performance budgets

Regression budgets enforced by `tests/perf_gate.rs`.
Run `cargo test -p mailrs-delivery-executor --test perf_gate` to check.
Run `cargo bench -p mailrs-delivery-executor` for criterion baselines.

## Path taxonomy

All public ops are sub-millisecond on M-series Mac. Budgets sit at
~10-50× the observed median to catch order-of-magnitude regressions
without flaking under cargo-test-workspace parallel CPU contention.

## Measured (release, M-series Mac)

Numbers populated by `scripts/stone-audit-footer.py` from criterion
runs — see README.md "Stone audit" footer for current snapshot.

## How to add a new gate

1. Pick an op that's on a real hot path
2. Add a `#[test]` with `time_median` measuring it
3. Pick budget at 15-30× over the observed P95
4. Document the budget here with reasoning