While we're speculating on the possible causes of the DNA 40's glitches etc, I wouldn't be surprised if the board layout or manufacturing tolerances have caused a track or tracks to be not substantial enough, causing unwanted voltage drop and messing with the power supply to the Atmel mcu, especially at higher power. Maybe just a contributing cause to the other possibilities previously mentioned.
However, I don't understand why Evolv didn't put some statements on their web page keeping people informed, it would have saved a lot of angst. I am also disappointed that Evolv does not put PCB version numbers on their PCB's, this is an industry standard practice, and allows owners or service technicians to properly identify a version of the board. The faulty boards in circulation now cannot be identified, ever, not even in 2 years time when someone sells some on ebay. Also the data sheet lacks detail, expected behaviour of the display items, and revision status of the document.
It also appears that the Beta testers were using 25W boards with temp protection, rather than the DNA 40 boards. So the DNA 40 board appears to have not been field tested, or if it was, by few testers.
I bought 2 boards, I thought Evolv were thorough from watching the P Busardo videos of factory tours. I liked the people there. Testing for worst possible conditions was mentioned too, making their products bomb-proof. The boards are not cheap and I was expecting made in USA quality, I actually thought the boards were made in-house, but I don't have that impression now. My impression now is of a very young company and I hope they will have a good, long, hard look at how they do things.