I installed 2.6 this evening and took it out to do a bit of close-in, low altitude test flying to see if my tuning was still good, just before it got dark here. Before today, I had been flying 2.5.5 without issues (besides a yet-to-tune loiter). This first flight with 2.6 went fine too, until...
...a few minutes into the flight, while I was in Stabilize mode I think, the quad suddenly started accelerating vertically, climbing away. I zeroed the throttle, and jumped into Land mode, but got no reaction. I was flying next to a tree, and the quad lodged itself high in the tree crown, preventing a total fly-away. It hung there, props spinning at high throttle like a weedwhacker, for a very-long feeling time (maybe a minute), until it came loose, tumbled down and smacked in the ground.
I flipped through my modes and wiggled the controls while the quad was in the tree, getting apparently no reaction, but I did not have the presence of mind to try to disarm it. (Note to self: remember to disarm.)
A couple of broken legs, a set of chipped props, a few popped screws and a broken cheap camera mount seem to be the only hardware crash damage - not a problem, thanks to the 3DR frame engineering. As a positive side effect, it was good to experience that my heart does not actually stop in this type of situation :-)
In my flights over the last few weeks since building the quad, I have not experienced any unpredictable behavior, scares or radio problems. The props were balanced and the battery had plenty of power.
3dr-b frame, 880kv motors, 11*4.7 APCs, APM2, 3dr radio, sonar, optical flow (not tested yet), 4s1p LiFe 4200mAh battery, DX8 TX, AR6210 RX with satellite, DSMX.
I am attaching the log. It gets interesting just after 90% / 600s.
Needless to say, correlation is not causation. This might not even be a 2.6 problem or (like many quad issues) a software problem at all.