Whats a safe TH go adjust to without burning boards out?..
Thanks
All miners have safety built-in cut-off temperature (can be disabled), where miner stops and cools down. Maximum speed depends on chip and VRM temperature. My own experience is that S9 survive temperatures over 110C at chips for 24/7 operation. For those who do not agree just looak at Antminer R4 which is "silent" and always works at 110-115C. Anyway, based on earlier S5, chips they unsoldered themselves under full load, so temperature exceeded 200C. They worked again when resoldered back.
So back to temperature: It is not possible to check VRM temperatures, but first batches of S9 was without heatsink on VRM. They worked well at standartd frequency, so newer batches should work well when OCed. My opinion is that only VRM temperature is the only limit of OC level, so keep temps as low as possible.
Unreliability of S9 leading to lost chips is caused by unpredicted behaviour of chips in serial chain. This will cause temporary overvoltage of some chips which will destruct them. Only one dead chip can make whole board stop hashing or prevent detecting of ASICs. Unfortunatelly, S9 still hasn't reliable overvoltage protection.
I do not agree you will shorten life of your miner in any measureable way. Chips are deeply under its maximum performance, only one unknown is VRM. Anyway I agree that even unmodified S9 is not reliable. Hope my opinion will help you.
FYI: Alexander1560 from the video claims he has working S9 at 850MHz.