You mentioned testing with the engine running - My guess is that the signals you’re looking for are much lower-frequency than the noise from the engine, so it seems to me that it would be useful to experiment with a few different mounting schemes to see what you can do to soak up vibrations. (In other words, implement low-pass filters in hardware.) RC helicopters use gyros and while engine vibration does impact performance, a little bit of foam helps quite a bit. The soft foam used for sealing gaps around windows worked well.