Please see my forum post below. Sorry, but I really don't want to have to repeat myself here.
Gist is that the huge backlog in snapshot replication (because of any number of reasons) continuously errors out after exceeding 'kern.ipc.maxpipekva'.
Something in the replication processing of a backlog is opening pipes and not closing them. A reboot of the server is the only was to clear the condition. But as replication resumes, it eventually returns to this same error condition.
Again, details are in my forum post. Thanks!