Didn't measure it. By the datasheet, the operation fequency is about 1kHz, and at least 10 cycles are required if C2 is  minimal, which would be 0.1µF. As C2 is 1µF, switching time might be about 0.1s. Used the circuit with a reflective surface; to avoid spurious switching, it is rather slow. But 100kHz frequency and thus 1ms switching time should be possible.

Interesting! For most applications that is probably plenty enough. I'm now looking for a problem which can be solved by your circuit :D

Used it for a telescope drive many years ago, sensing a white area as within limits. Too much ambient light as well as a broken or not connected wire would stop the drive.  

Very nice idea. PLL's have gone somewhat out of fashion. I guess the performance depends on the LED, photo transistor, and thePLL-frequency. Do you have any practical data on the performance in terms of how fast it detects with the components you've used?