On-premises data gateway-Network Request Returned Unexpected Error-Configuration client returned nul

You'll want to distinguish between continuity and recovery.

For continuity purposes you absolutely will want to have more than one cluster member.

For recovery purposes it is critically important that you do not remove a broken anchor member from a gateway until AFTER you have recovered its function with another cluster member.  Failure to do that will result in loss of all gateway connection details that are stored in the Azure cloud.

In short - you can recover much quicker if you have a spare VM ready to go at a minute's notice to take over a broken anchor member. Or you can avoid the downtime altogether (in most cases) when the other members take over.  (There are corner cases where the entire gateway will not work until the anchor member's management console has been logged into)