Suppose we are doing state machine replication with voting, allowing for up to F process failures.

How many replicas do we require to achieve a write quorum despite failures?

Suppose the coordinator $($for an update$)$ and several replicas crash. $$What guarantees that at least one of the surviving replicas will know of the update?

Suppose two clients perform updates U$1$ and U$2$ concurrently on the same replicas. $$What guarantees that U$1$ and U$2$ will be delivered in the same order at all replicas?

Suppose a replica is down when a client reports the result of voting.$$ Why is it important that the replica $($after recovery$)$ find out what the result of the vote was?