@Marie Horton yeah the dynamic range switching approach is exactly what I'd recommend for your use case — the tricky part is tuning your threshold so you're not flip-flopping between ranges constantly during borderline vibration, which can create noise in your data. A simple hysteresis band around your switch point helps a lot there.
One thing worth thinking about for early-onset resonance detection specifically: even at ±16g you'll have enough resolution to catch frequency patterns, but if you're doing FFT analysis on the data you'll want to make sure your sampling rate is high enough to capture the resonance frequencies you care about before they climb. The sensors module runs the accel alongside gyro and magnetometer so just be mindful of how you're scheduling your reads if you're pushing high sample rates.
The AI Firmware Builder in the Dev Hub is actually pretty decent for scaffolding this kind of conditional range-switching logic — describe your threshold behavior and it'll get you 80% of the way there without having to wrestle the register map from scratch. Still worth cross-referencing the datasheet for the exact sensitivity values per range though.
