Add a lock for persisting service/check state to avoid a race

Right now the persistService/persistCheck functions can be called concurrently for the same service and attempt to update the same file on disk at the same time, resulting in undefined behavior. If we use a lock, we can make sure the later update is persisted.

Start by locating the persistService and persistCheck functions and examining how they update their on-disk state files. Reproduce or inspect concurrent calls for the same service, then verify that updates are serialized and the later update is persisted without undefined file behavior.

Written by the indexing model from the issue text.