TSAN reports a data race in the lock manager here and here.

To fix this, all usages of the lock request status should be protected by either the bucket mutex or the partition mutex.