Major Description
The behaviour we saw was that considerable time elapsed between
different cores within the same solr instance publishing themselves as down.
Separately it appears from the code that some cores would not be published as down if another core returns from the function early because it will be its
shard leader (return vs. continue in for loop).