Azure Autobalanced Collectors - 5 minute windows of ping failures

I can't find the discussion but I recall there was some issue brought up before about the way that Java/LM does ping would not not increase some packet id # which would cause some routers/firewalls to block them eventually, and restarting the Collector would clear it. There are some situations where I've had to restart a collector to fix a ping issue. But it's always been that ping stops working completely until the collector restart and not just stop for a few minutes then fix itself.

Is there anything special about the routing between the collector and a device that stops pinging? Does it happen even if you are pining 127.0.0.1 (like pinging the collector itself)? Are the devices affected all on the same region/network/subnet? Is special routing table or 3rd party router/firewall between them? Perhaps worth running Wireshark or network watcher during the issue (if you can catch it live) and see what the traffic looks like. What does the LM wrapper/sbproxy log show if anything? etc.