1) Any time I see nvarchar(4000) parameters I think ORM, mismatched data types, and a butt-ton of CONVERT_IMPLICITs with their correspondingly horrible performance and concurrency issues.

2) You clearly have repeated complete buffer flushes. That's brutal.

3) Solarwinds is aggregating all IO stalls into 15 minute buckets. Your graphs just show latency, not how many things were running concurrently and collecting fileIO stalls. Just yesterday I captured a 180 SECOND run of IO stalls on a client and they totaled a mind-blowing 2.5 MILLION MILLISECONDS in duration!!! That is 28.94 DAYS of IO stalls in 180 clock seconds!! Guess what, you can't run almost 2TB of BUSY and REALLY BADLY DESIGNED AND CODED OLTP databases with 2 synchronous AGs for each database on a four 6-disk RAID10 sets. :hehe: