How is replica instance failure dealt with in AWS Aurora?

Question

The AWS documentation explains the failover mechanism in the case of primary instance failure, where one of the replica instances is promoted to be the new writer instance and the old writer instance becomes a reader instance. But there is little to no information about how reader instance failure is handled.

Does the reader endpoint point temporarily to the writer instance while the reader instance is being revived and later switch to the reader instance? Or does it simply fail connections till it's brought back up?

I need to know this to decide how long I can try reconnecting to the reader endpoint. If it redirects to the primary instance, I'm guessing it'll be faster. Please do help. Thanks!