Performance Issues with a Self-Managed PSSSA Replica Set

Ops and Admin Replication
1

Reduced persistence when saving documents is known for the PSA Replica Set setup and the failure of a data-bearing node as described here:
https://www.mongodb.com/docs/v7.0/tutorial/mitigate-psa-performance-issues/#std-label-performance-issues-psa

However, we also observe this for a PSSSA setup.

For our PSSSA setup, we have set the default write concern to w:2 in order to be able to cope with the failure of two data-bearing nodes. Unfortunately, we realise that the write performance drops dramatically if two data-bearing nodes fail in this setup. This is not documented for a PSSSA setup and I would not have expected it.

Is there a setting that changes the behaviour? Even with a write concern of w:1, we unfortunately observe the drop in performance.

I tested with MongoDB Community Server 7.0.15 and 8.0.10.

Thanks for any hint in advance.