The concurrent access occurred in the region merge/split logic, as it did not own the ticket lock. To resolve this, we simply make the ticket add/remove acquire the read lock of the region lock.
