Swarm.Tracker stuck in `syncing` state when lots of nodes starting/crashing

[darrenclark]

I can't reproduce this on reliably, but I've noticed that occasionally Swarm.Tracker gets stuck in the syncing state when a lot of nodes are crashing & restarting.

Based on inspecting the process on each surviving node last time I saw it, it looks like they are stuck in the syncing state, and have a list of pending_sync_requests, including PIDs from nodes no longer in the cluster. (I think)

[lATAl]

Same issue. Can't reproduce but when I call :sys.get_state(Swarm.Tracker), some node return

** (exit) exited in: :sys.get_state(Swarm.Tracker)
    ** (EXIT) time out
    (stdlib) sys.erl:315: :sys.send_system_msg/2
    (stdlib) sys.erl:116: :sys.get_state/1