nodeCategoryChanged event on already-down node makes extra nodeDown events

Steps to reproduce:

1. Create a requisition with a single, reachable node in it and synchronize
2. Wait for the node to be polled once successfully
3. Remove the polled node from the network so that it enters a node-level outage
   1. A single nodeDown event is created when the outage is opened
4. Add a category to the node in the requisition
5. With the node still down, synchronize the requisition

Expected result: A single nodeCategoryMembershipChanged event is created for the node
Actual result: In addition to the expected category-related event, a handful of spurious nodeDown events follow a short time later