Time to time work with electronics is deeply confusing.
You may have 3 identical FET drivers connected same way, but still - only specific one burns out. Again and again. I've changed resistance (Rgate), and it seems that situation with heating improved a lot (switching now occurs in a neat way, without confusing oscillations, I was impressed: originally thought that high control currents would make everything better, but it seems that they require a bit more thoughtful layout with damping circuitry)

However, one transistor from six still overheats, low-side. Reason is simple:

It never got sufficient voltage to fully open. I have no idea why one of sides died this time, or maybe it's components around or flux remained on a PCB? Hope I would find a reason. If not - I will make new PCB for trials with dip panels for drivers to change them without tedious unsoldering procedures.

While producing prototype of CDI (maybe I would release it as OHW at some point) I found my personal favourite way to manufacture PCBs at home. Previously I tried to mill them entirely, however it's time consuming if you want to remove a lot of cooper (and you usually do), it seems that with a cheap router you can get much greater result by combining etching and milling holes + board outline