I don't have any solutions yet but I'll share some of my findings anyways.
Config is 2x prisma's, 3x 1000w PSU and BE200 controller running v 5.47, 12v 2a power supply for controller.
layout: BE200 controller - 1/17 - 2/18 - 3/19 - 4/20 - 5/21 - 6/22 - 7/23 - 8/24
all screws are tight, missing screws replaced, fans pulling air towards heatsink, handles facing inwards
air intake temp 50-80f
Slush's pool seems to auto generate a worker name which blocks manually setting difficulty. I think the auto difficulty is causing the hashrate to drop to as low as 200GH/s if left alone overnight.
BTC Guild worker names work properly, >/=2048 or =256 difficulty causes low hashrate
The biggest issue I see is the controller rebooting randomly and not detecting all boards. This happens with 3 different controller power supplies, clock speeds between 220-250MHz, 50f intake temp, controller connected to board 8/24, random board orders or either prisma individually. I started keeping a log of what boards are detected after rebooting to look for patterns and the weird thing is sometimes it sees just 1/2 of a board towards the end of the chain. For instance, it saw just board 23 after one reboot which is towards the end of the chain.
1)run one at freq 220.
2) turn handles out
3) add a good fan push pull
4)set your btc Guild worker at 1024 or 512
5) use a 5volt 1 amp power supply for your controller