As I've seen this particular issue manifest itself several times, I just wanted to offer any information I have on this:
This particular alarm is monitoring the perfmon metric Hyper-V Virtual Storage Device > Error Count on the host server housing the VM (with each instance of this counter relating to a VM running on the host). From what I've seen myself, powering off or migrating the VMs and then rebooting the host server makes the error count reset, however, you can usually observe the error count in perfmon climb right back to where it was previously as soon as the host and VMs are powered back on so it seems that, in at least most cases, the commands causing the errors are happening during the boot sequence but I have no way of confirming this.
Unfortunately, as you've all noted in this thread, the environmental details surrounding this are often quite random. I've seen it happen with both SCSI and IDE disk controllers. The only strict correlation I've noticed is that this always seems to happen with Windows Server 2012+ or Windows 8+ VM's that are running on Hyper-V 2012+ hosts. I hope this helps in some way but I've not found any information out there yet documenting a cause or a resolution.