What went wrong?
The main problem was apparent loss of radio control. The rig worked fine at ground level, but behaved arbitrarily when airborne.
Specifically:
The shutter release was unreliable. The servo bears directly (through a cam) on the shutter release button, but less than satisfactorily. It seemed at times to be behaving as if the radio was out of range (although this was never confirmed, and seems unlikely.)
The tilt servo seemed to work O.K.
The pan servo, however, was definitely being asked to do more than it could handle. The servos operate through 90°—but a full 360° is required to pan the cradle. When geared up x4 to achieve this, the mechanical disdvantage was excessive. At ground level the servo would spin the whole cradle through 360°, but as soon as it was subject to wind pressure at altitude, it seized up.

Hence I could not aim the camera.

Additionally, the camera is a bit heavy and it has a dodgey auto-focus system which is easily fooled.

These shortcomings will be resolved in MkII by using a new lighter camera. (Canon Snappy LXII) It has an electronic shutter release, fixed focus and is lighter than the 35AF it replaces.
The pan servo problem will be resolved through a reduced gearing system described on another page.