I think ultimately you are perhaps concentrating on the issue of 2 or more partitions existing for an extended period of time.
You must be referring to monsterer's comments. I was always emphasizing that 2 or more temporary partitions can be tolerated if there is a resolution mechanism.
If the partition duration is short, the more likely it is that you can merge and resolve P with no data loss. It is my opinion based on fact and research that blockless data structures can achieve better results and less data loss than block based ones.
I think only PoS or PoW with blocks can be that mechanism. I will have to wait for details of your protocol to point out why.