In general, it is best to locate the Coordinators as close as possible to the SQL instance where Change Auditor database resides. This will help ensure reliable bandwidth and low latency when dealing with Coordinator/SQL transactions.
This is of course is rule which has exceptions such as multi-forest deployments, DMZ requirements, etc. where based on product architecture Coordinators may need to be deployed remotely to SQL.
If running Coordinator in remote scenario with high latency to SQL consider the following Coordinator SQL timeout registry keys if frequent timeout warnings are encountered in the Coordinator service log file:
SQL Timeout (general timeouts)
Bulk SQL Timeout (event insertion)
The suggested value to start for both registry key values is 300 (5 minutes).