colliding checkpoints with pull-only and a push-only replication

Found in a forum thread: If you start a pull-only replicator and a push-only replicator, with otherwise the same settings, they end up with the same checkpoint docID. That means they'll create conflicts saving their checkpoints on SG.

The visible symptom is probably that replications keep starting over from zero.

Workaround is to create a single push-pull replicator instead, but that may not always be feasible. (For example, you might need to track when the push finishes separately from the pull finishing.)