I was getting nervous with one peer of the group down so I put the result of the bulkload p directory for that group back in the place of the corrupted p directory, and the leader sent a snapshot to that peer - I think thats ok…
This does not solve the real issue here but maybe helped my cluster become normal. Or, I have messed everything up and do not know it.
edit: ok that was a horrible idea, that server was then missing data after it came back and looked healthy, I think it was missing all txns from bulkload->when it corrupted. Meaning the snapshot only applied after the corruption to current.