Amazon Aurora V1 inaccessible after scaling for long periods

Sometimes after a scaling operation on Aurora Serverless v1 (mySQL) and the database becomes completely inaccessible for a considerable amount of time. Today the database was not available for 2 hours (i.e. accepting connections) until it recovered.

Sometimes recovery can be manually triggered by a "force-scale" up or down in the console.

We had to created a new Aurora Serverless v1 from an snapshot as Point in Time recovery was not available either. The new cluster was twice unavailable again for about 2 minutes with messages like

The DB cluster has scaled from 8 capacity units to 4 capacity units, but scaling wasn't seamless for this reason: An internal error occurred.

or

The DB cluster has scaled from 4 capacity units to 8 capacity units, but scaling wasn't seamless for this reason: An internal error occurred.

It is set to roll back the capacity change if a scaling point cannot be found and even if connections were dropped, I would expect being able to re-connect, but not being locked out without any options for periods of time. Since it's serverless, we don't have any insights or even the ability to restart and the message "An internal error occurred" is rather vague and unhelpful.

How can this downtime be prevented?