Thank you for reporting back - did performance improve as well as the crash frequency being reduced?
Why would you need a 200GB tempdb? Index creates and/or rebuilds on 86.2GB-1.2TB tables, of course!*
Check your metrics on the SAN - is it heavily loaded in terms of IOPS, in terms of throughput, or both? If it's IOPS, you've done well to get new spindles also on the SAN. If it's throughput, then a better solution is to provide enough IOPS and space locally, which will give you the tempdb performance you need, as well as free up SAN throughput for the rest of your (and other) uses.
Local SSD's in RAID1 or RAID5 are excellent for tempdb use with modern controllers - and yes, if you benchmark them, you'll see that RAID5 performance and RAID10 performance are pretty close, and if you're using that few spindle disks, either one with a handful of good SATA or SAS SSD's (3 100GB SSD's in RAID5, say) should trounce whatever performance you're getting now.