Multi Primary vs Single Primary Architecture for power DNS Authoritative Server

[Yashank-R-S]

What are the considerations to be taken into account while deciding implementation of the following self-hosted DNS Authoritative Server Architecture:

1. Multi Primary Architecture: 2 Master pDNS servers and 1 Slave pDNS server. This gives me the advantage of not having to handle master pDNS failover because 2 servers are available for writes.
2. Single Primary Architecture: 1 Master pDNS server and 2 Slave pDNS servers.

FYI: There is dns-dist proxying DNS Queries to pDNS before both the architectures.

What considerations should be taken into account to decide between these 2 architectures?
Are there any downsides of Multi Primary Architecture, while it seems most reliable?