It makes the SITL sensors behave much more like real sensors by creating a separate instance for each gyro/accel pair. With this I can write a perfect FFT test for my FFT PR, without it it breaks all sorts of assumption in the INS backend code. The code is also simpler done this way. (it does not create a thread which I know tridge didn’t want).
It fixes the SIM_VIB_FREQ simulation to use a consistent time period, without which you get all sorts of random harmonics in an FFT
It stops the backend rate being set to the sensor rate. None of our real sensors do this and doing it again breaks assumptions in various places. You can still do sensor rate logging because we simulate that correctly now.