I measured four Skywalker 40A ESCs (Silabs processor) at room temperature, and the standard deviation was 4.4kHz with a mean of 24.018MHz, this amounts to a std deviation of 180ppm (or 0.018%, 1.8e-4) which is in the range for a poor crystal, better than most ceramic resonators.
The range was 10kHz which is 416ppm or 0.04%.
Note though that the specs on the Silabs C8051F330/1/2/3/4/5 oscillator at factory defaults are 24.0-25.0MHz, that is +/-2% or equivalent to +/-20µs servo input. The power on OSCICL value is factory calibrated for 24.5MHz.
At the time of the test, BLHeli decreased the osc frequency to 24MHz though the inline comments suggest it increased 25MHz, but the author seems to have misunderstood the OSCICL register.
I always do throttle cal on Silabs based ESCs.