inconsistency issue caused by lock failover during node joining

Scenario:
tx1 writes to k, hash(k) =

. Tx is prepared, commit hasn't been send yet.
New node C has joined so that ch(k) =

, tx1 hasn't completed.
tx2 writes on k adn prepares/acquire lock on k@C.
both transactions commit at the same time resulting in B seeing tx2 result and B seeing tx1 result.