Waring: v2.0 is not reliable

I've put a warning on the download page...

I can make it to reboot while configuring: for example keep changing the tx power quickly up and down, and it will restart within one or two rounds, but it flies without issue for me too...

FDR wrote: I can make it to reboot while configuring: for example keep changing the tx power quickly up and down, and it will restart within one or two rounds, but it flies without issue for me too...

I've increased the watchdog timer to 2000ms from 1000ms, and while it was much harder to make it restart, it still happened...

The restart is almost certainly started by the watchdog, but it shouldn't happen. It means there was no event loop or screen refresh in a whole second. That's a sign of some error in the code...

Prolonging the watchdog interval gives more tolerance to a possibly long process, but cannot solve an infinite loop...

The big change for 2.0 (with regards to stability) is the movement of the mixer evaluation to the interrupt handler. While it is also possible that telemetry is causing the issues, it seems unlikely to me mainly due to there being no loops in that code. If the model were losing connection to the tx, that would be more likely to be a telemetry issue I think.
Interrupts can be tricky things. you need to be sure that no place do you depend on a value not having changed while you are working on it. Effectively it is just like threaded programming.

In this case we actually have 3 different interrupt levels:
level 1: the mixer
level 2: the radio
level 3: the msec increment

level 1 interrupts the main loop, level 2 interrupts level 1 and the main loop, and level 3 interrupts level 1 and 2 and the main loop.

level 1 is new to Version 2.0.
Also, the main loop (where the watchdog is) is now triggered by level 1 as opposed to running constantly.

If the time it takes to process level 1 + 2 + 3 is greater than average time between interrupts, than the main loop will never fire and you'll get a reboot.
Note that the time to process level 1 is related to the number of available mixers, though it should still not be a bottleneck. The time itself is deterministic, but the overlap of the 3 interrupts is not.
Every transmitter should behave exactly the same way, so it would be helpful to get model files that cause the issue.

You can reproduce the reboot with constantly changing the tx power even in the defult empty model config...

Probable, but I'm not sure about that.
Reboot happens with simple mixer tweaking too, it's just easier to reproduce with the tx power.
The cause might pretty much be the same....

No, it's definitely in the fw...
The model.ini is a plain text file.

Obviously this is my top priority now. I will spend some time trying to root-cause