So your idea is to simply fail everything? Presumably one of these is a production database - one where the "work" actually happens and one that is needed for "real work" to actually be done. And, presumably again, the other two act as some sort of backup. Disaster recovery perhaps, reporting maybe?
"Throw your hands up and surrender" does not seem to be an appropriate way to handle replication issues in a production system - at least to me.