One feature of most fan headers is monitoring the fan for failure - that is, for a speed reading that is zero, or less than some trigger value. SOME mobos actually allow a fan to stop at low temps, but I think that's not a good idea. Anyway, assuming that stopping is NOT considered normal, all mobos take corrective action when a fan stops. When that failure is detected, the header raises the fan signals to call for full speed so it DOES start up reliably. If the fan does re-start immediately, the header then reduces the speed signal to where it was before. If the fan fails to re-start, you get a fan failure warning on screen.
If that process were not in place, the alternative would be this. The fan would stall when run too slow, but there would still be a small current flowing through it because there IS a voltage supplied, just not enough to re-start the fan. That heats up the motor, but it is not turning so it cannot cool itself. Moreover, the item it is supposed to cool (in this case, the CPU) would be getting NO cooling air flow, so it would heat up. As it does that, the mobo header would respond by raising the speed signal to the fan in response to higher temps. Eventually the signal would be enough to re-start the stalled fan, which would restore cooling air flow until it cooled down again. And repeat. In the meantime, each cycle like this would produce a heat-up period for both the fan and the target heat source, and then a cool-down period. Thermal cycling like that is to be avoided; hence the built-in process to detect fan stalling and try to re-start immediately.
Probably you are observing this process repeatedly. The fan is re-started and returned to its slow speed at a setting that it too slow, so it stalls again, and repeats the re-start-slow-down-stall cycle. As DRagor said, change your minimum speed setting to a higher value.