The synchronous volume replication goes in this way:

1. At the first connection or reconnection the replication compares ECC tables from source and destination. Those parts that have different checksums are transfered first speed limited by speed set in replication task or by speed of the connection. Until all the changes will be send to destination it is stated as INCONSISTENT. This part works like an asynchronous replication.

2. All write access in the synchronous replication action are:
- buffered by source system,
- send to destination system,
- written to disk by destination system,
- confirmed by destination that they are on disk,
- written by source system to disk.
After all those steps next write access is taken.

When connection is dropped the replication task is stopped and no replication is running. The data are written directly to the source disk and ECC checksums are changing. When replication get the connection again the first step starts again.

This shows that whatever - source or destination server is slow, network connection is slow - the speed will be limited by the slowest part of the replication servers. You need to have fast network and fast raid on both sides to have successful working volume replication in failover.

Please note that the async volume replication don't do the 2 point and the replication works all the time more like the 1 point only. In result that gives no guaranty that when the source server will fail the destination server will have consistent data. This is the reason why failover requires synchronous replication.