Your query looks good for finding problem reports. Adding in the byte count was a good touch.

I'm not an expert on wait types, but I think you've got the gist of it. My understanding is PAGEIOLATCH_EX waits are involved in pulling data from the disk into memory. High values often mean either contention (which it sounds like is the case for you), or disk bottlenecks.

To deal with contention, one possible option might be to stagger the subscriptions, so instead of all firing off at 2:00, some go at 2:05, some at 2:10. etc. That way some of the the IO load is spread out over time. Another is to identify the worst performing queries and tune them so that they aren't in contention for as long as they are now. Are any pulling up more data than they need and then filtering it down, or are there missing indexes that could speed up reading data off the disks? Those are some of the options I'd be looking at.