I'm curious how wide the failure window was (timespan, ramp-up/down, etc), relative to how many devices were involved. And I wonder how well the signal in that

I'm curious how wide the failure window was (timespan, ramp-up/down, etc), relative to how many devices were involved. And I wonder how well the signal in that ratio might scale down to hundreds or tens of disks.