It's been a while... I have an update, but not a resolution.
So I finally took ASIC to maintenance.. after few hours of diagnostic - the resolution was 18 burn chips in total on 2 plates.
Well.. it is was it is, got it fixed ( not cheap ofc ), after repair asic was turned on without any issues, 40Th - all good.
Took it back to "farm", plugged in, did autotune, all good. Started mining - for one hour it stayed 40Th, then dropped to ~25.

In the morning checked the logs -
Hashplate [1] - 25/30 chips

After multiple restarts, change of profiles  -
Hashplate 0 - 0/30 chips
Hashplate 1 - 25/30 chips

So obviously, again chips are burned... but since it worked only for an hour, its pretty clear that chips are an only consequence of something else failing on ASIC.
There are not many options, either control board or PSU, but I really have no idea, and after first repair I'm not sure how much I can trust repair shop...
Any comments?

Ok.. so update on this topic.
First of all, I was wrong that it's always a random hash plate - it is always chain[1].
1) On default firmware nothing changes
2) On chain[1] plate one capacitor for the power supply was a bit loose, soldered it, no difference tho.
3) Taking a functioning hash board and placing it in a chain[1] slot does not solve the problem - still in the log it is always chain[1] that is falling off.
4) Took ribbon cable from functioning board and placed it in chain[1] - no difference.

Took everything apart and inspected, no visual damage, all heatsinks are sitting tight. plates look like new.

Again - can faulty control board or power supply cause such symptoms?

I appreciate your time and help, but it seems that you are focusing on one hashplate.
I’m not an expert at all, and I don’t think that it makes sense to talk about reasons for flashing etc.
I’m not hoping that there is some simple trick to make it work again. If it has to be repaired then so be it.

What is strange to me that every time it is different hashplate that drops off. I hardly believe that suddenly all plates are bad, and there was no physical damage to unit.
Therefore, my main questions are can these be a symptoms for bad control board or PSU?

First thing first I will try to flash default bitmain firmware on it.

I’m this kernel log it is chain 1, after reboot it could be chain3 and so on.
If one bad plate affects whole circuit and can trigger different hashplates that’s a different story.

If it always would fail on one hadhplate I would not ask, but plates change.
This boot it was plate1. After restart plate 1 might work well, but plate 3 would not be visible. After restart plate 2 and so on.
It was running for 30 minutes on default OS with stable 42Th, but I don’t think that it really matters as 30 minutes is nothing

After flashing hiveon OS on T17 it started to randomly drop hashplates as in attached kernel log. Sometimes it works for 10 minutes, sometimes happens immediately. And it’s basically always different hashplate.
I can’t say that it worked fine with default firmware, as we did not test it.
Any ideas what could be wrong?
https://pastebin.com/saQGcNxi