Check distributed transactions

It might be that distributed transactions' "commit" op lends with different cluster times on different shards. Hence after the restore, the same txn might end up committed on some shards and "unfinished" on others.

Need to check if it's true, assess consequences, and fix if needed.