I have been playing with these cameras for some time using opencv/raspicam to control the camera. The raspicam/opencv interface allows setting the gain, this is dubious because there is no manual set of the gain parameter exposed via V4L2.
When testing the camera (even the older V2 sensor) while using the opencv/raspicam interface the auto exposure algorithm seems to take back control. With this in mind I decided to get rid of the middle man and use the V4L2 drivers.
V4L2 seems to work fine unless you start setting initial parameters when the camera is transmitting frames. I set the video_mode values to 4 and in the receive frame loop I reset it to a different integer (11 or sports).
This caused the camera controls to become unresponsive. I restarted my application taking out the video_mode set in the receive frame loop.
Once the software restarted the camera controls would not work and the camera appears to be stuck in an unresponsive loop of some type. The only way to rectify this problem was to power cycle the Pi and camera.
I'm wondering if anyone has seen this behaviour before? (I'm confident I can avoid setting initial parameters in the receive loop but felt the condition was worth a post on here)