Update
It only looked like at first there is some sort of protection inside AR9271. Turns out there isn't.
Part of the chip got fried, there was a consumption of about 0.5W from wall and chip getting slighthly warm.
Even in such state it was still useful.
But after having problems with RT8020 replacement and one more dead power converter i managed to solder another RT8020 with good connections.
Plug in card to laptop, it works. Plug in to router it works in both AP/STA modes.
But when i tried to load it on my modified router firmware, STA mode kept crashing no matter what.
Then, after some debugging it was discovered it works fine in AP mode accross entire range 2192-2887Mhz, however, when changing channel from approx. above 2550 Mhz to lower channel it would crash.
Same way it was crashing in STA mode, because in STA mode the card first scans entire range and then goes to selected AP channel. To confirm i set another AP at 2712 Mhz and the Alfa in STA mode connected without issues.
Then it was also noticed the TX/RX rates are not full with alfa on 2.4channel in AP mode and laptop connected to it. There was RX 108mbps and TX mostly 150mbps but dropping to 135 and 121mbps.
Next came the idea to warm the card with 150 celsius, after just 20 seconds, both rates synced at 150mbps stable. It cooled down, rates dropped.
Heated again, this time directly on AR9271, rates good again, tried changing channel to 2700mhz and back to 2412mhz while chip is hot and there was no crash.
I thought maybe solder got ruined from all the heat during RT8020 replacement attempts and tried to resolder all the elements and ar9271 contacts.
After this the card kept working and switching channels to and from high band worked without crashes.
But i noticed the txpower was very weak. One or more pins on SE2576 amp lost contact, i fixed it and the power was good again.
But the crashes came back as well.
At this point i was in doubt on what to do next:
Seems like heat on AR9271 restores everything to work fine, so it could be damage on it.
But then again, with SE2576 disabled switching channels worked fine.
I had TL-WN722 with same chip. Took off both AR9271, and put the one from tplink to alfa.
The problem with rates and channel switching remained. Ordered SE2576 parts and waiting now to arrive.
If replacing that doesn't help then it will probably be the most annoying capacitors/inductors mess...
The resistances of 3.3/1.2 power paths on alfa were about 159/143 ohm. After too much heat got to it during RT8020 changing they dropped down to 28/12 ohm. And during further repair attempts they rose to 60/20 ohm.
This is all bad. With good chip from wn722 on alfa, i get about 1kohm on 1.2v path and climbing up to 230kohm on 3.3v path. This may not be final good values depending on type of failure of SE2576 or capacitors.
The consumption from wall of the alfa card is 0W after ar9271 changed.
The chip taken off alfa had only 17ohm on 1.2v. Since it was still working i decided to put it on tplink.
After soldering, it had 16ohm/46ohm. WN722 works with chip from alfa, even though it is partially fried and having such a low resistances, TX/RX rates synced at 150mbps and kept stable, also no crashes when changing channels accross 2.2-2.9ghz range on my customized firmware.
So, the chip seems fully usable, with small dowside it consumes some extra power and generating a bit more heat.