The first time ckpool just stopped and dmesg said it segfaulted but alas no core, the others were lock timeouts.
The KDB restarts were both lock timeouts and me restarting it to work around the resource issues by changing things.

I've found recently that ckpool is a CPU hog and I'm pretty sure I know why
(it uses between 2 and 3 times the CPU of KDB but does about the same amount of work)

However, regarding the resource issues, if it happens again in the near future I'll expedite upgrading the main back end server.
It should be enough to handle the load, but this example says otherwise when unexpected things occur.

Of course, no one loses anything, but I guess having to mine for a while on your failover pool can be a let down
(I certainly don't want that to happen to everyone)