Okay, finally gave my ArduCopter a try today. Freshly charged battery, nothing fancy, running ArduPirates code r527.
I've observed two major issues:
1. the yaw drift isn't really getting corrected by the magneto AT ALL - it kept drifting, usually clockwise (not sure if that's significant)
2. the y axis (i.e. altitude) oscillations made anything but short hovers (second or two, perhaps) completely impossible - i'd have it hovering stable and then it'd suddenly start dropping like crazy, touch the ground, then i'd overcompensate and it would shoot in the air to about 2-3 meters, then i'd overcompensate again and, as soon as i'd finally get it to hover, a second or two later, it would drop again. It looked like it was on a trampoline, prompting a random bystander to comment with: "That a prototype, or something?"
Re: 1 - Yes, the magneto does work, is enabled, the offset has been calibrated and its north indeed is, roughly, real north. The center stick position does correspond to center in Configurator.
Re: 2 - It was somewhat windy, so that might've attributed to it, but i doubt that's the sole reason for it. The oscillations were pretty regular, for one. Also, the wind was nowhere strong enough to cause any significant slewing or oscillations along other axes. Now here's the interesting bit - enabling altitude hold (barometer only, no sonar yet) kept it in one spot for a bit longer than a few seconds, with audible RPM oscillations. Then it went out of range and started oscillating just as badly as when i was handling it manually.
Finally, one observation re: propellers. The way jDrones/DIYdrones kit propellers are attached to the motor shafts makes for a REALLY bad design. Why? Because two of the motors (front and rear) rotate in such a direction (clockwise) that any torque exerted on the spinning propeller UNSCREWS the retainer screw/nut.
Net result - i've had rear propeller fly off the quad after it hit dirt during a hard landing, which was lucky (it spared the propeller); however, i've also had the front propeller unscrew itself mid-flight (good thing i was keeping the quad no higher than 2 m at all times) and catapult the nut and screw some 5 meters away.
And judging by the sound of the impact against a building wall and an iron fence, they make quite a nasty projectile weapon.
Maybe it's just me, but shouldn't pusher props also have reverse threads on their hub nuts/screws?!