So lessons learnt - the bigger the rigs, the bigger the issues in PSU requirements, cooling strategy, cabling, stabilizing them.... a golden rule : stability, cool temps and consistent submission of shares to the pool's stratum server are more important than achieving the meanest powerful, fastest rigs that requires constant babysitting. If you have a large farm -- undervolting is very very key - forget the world GPU speed record.
Yes, you end up losing more in unexpected downtime and troubleshooting. Better to be cool and steady than fast and hot. That's good advice even for those not running a large farm. I learned that with my SP20s.