tapdisk 100% / guest VM high iowait

This problem exists on two production clusters, one running XenServer 6.5 SP2 and one running XenServer 7 with guest OS CentOS 6 and 7 respectively.

Problem presents itself as follows:

dom0: tapdisk process consuming 100% CPU. The process is not present in tap-ctl. It is sourced to the problematic VM by scraping through /proc

Guest VM: immediately begins experiencing high iowait times.

Remedy: Reboot of VM.

Important to note that the VM reboot does not clear the tapdisk process in question, it must be manually killed.

An strace of the tapdisk process yields a never ending:
gettimeofday(

, NULL) = 0

As mentioned, this is impacting two production clusters running XS 6.5SP2 and 7, where hardware and guest OS versions differ. The only similarity is the application being run on all guests (RabbitMQ).

We can provide XS service reports as required.