There are three improvements we could make to how the coordinator shard processes votes and sends commit/abort:

1) Use one of the Grid's "arbitrary executors" rather than the Grid's "fixed executor". Currently, the fixed executor is used because it is the only executor for which the a mock network is supported by the sharding unit test fixtures.

2) If a voteCommit or voteAbort triggers making the decision, do not block the voteCommit/voteAbort Client thread on things like persisting the decision, waiting for majority writeConcern, sending the decision and waiting for acks.

3) Make sending an abort decision a "fire and forget" message. Currently, the ARS does not support sending "fire and forget" messages.