Oh forgot to leave out that small factor, almost no clustered indexes (yes I know bad) but that is the one change if I make lose my warranty, trust me been a dba for 20 years I believe this company is developers gone wild). I did run cystal disk mark against 2 of my servers on the index drives (yes all my indexes are on own filegroup). One that the active/passive cluster that this database is on, and the other against a little newer server that is attached to the same SAN. Going by ozar's description for my heavy OLTP server the 4k QD32 is the most important.

for my problem server I get

22.8 MB sec read and 78.52 MB sec write

for my other server

29.62 MB sec read and 106.2 MB sec write

SOme difference but also guessing newer server also has a little faster controller.

Checking through the log table I found 5 indexes on 138 million row table took 4.8 hours to rebuild. In fact out of the 25 hours it took 30% of the indexes rebuilt accounted for 80% of the time