Yeah, that's "Instant PWM" kicking in.
the update rate are the same as the loop speed (each cycle, one update)
so you get a signal every run
as we all see, yes, they are :D
(Having looked at the implementation in APM_RC.cpp) Why isn't the timer/counter's 'TOP' value just reduced to 5000 rather than having that 'Force_Out0_Out1' jazz? Manipulating the timer/counter register to 'skip' over part of the timing seems unnecessary and terrifying.