The data service must be crash tolerant. That is, it must crash-recover disk data (if necessary) when it is started as the result of a cluster reconfiguration, scha_control giveover, or scswitch switchover. Crash tolerance is a prerequisite for making a data service highly available because crash recovery (the ability to crash-recover the disk and restart the data service) is a data integrity issue.
The data service is not required to be able to recover connections.