Yes if it detects only 28 ASICs then there is a big possibility that the chip is broken or near parts like shorted capacitors.
Regarding thermal adhesive for a loose heatsink, Artic silver is fine.
Chain [ 0] and [1] are broken based on your kernel logs running them one by one if it only detects a few Asic chips then it's a hardware issue.
I have already given you a guide above there are 13 voltage domains and each line has 5 ASIC chips if one chip is broken of any of these domain samples like in 8 one of the chips under that is broken and the rest 9, 10, 11, 12, and 13 won't be able to detect then the hashboard will automatically shut down because it didn't detect the required amount of ASIC. You need to replace it in order to fix the issue sometimes capacitors near that ASIC are shorted removing it is fine.
About the PSU if you already test them one by one but the result is the same then it's not the PSU the two hashboard is totally broken and it needs to repair.
Even if you change it to official firmware I believe the result will be the same. Unless there is some modified firmware out there that can able to ignore missing chips and only the remaining chips will run. Old units like Antminer s9 can able to ignore those bad ASIC chips and run like normal with lesser hashrate.