Static gyro (or accel) bias error is easily removed. Full-scale error, cross axis error, axis misalignment errors are a tad harder to correct but can be done also in a straightforward albeit somewhat time-consuming manner. Gyro drift is the main remaining culprit in 6 DoF dead reckoning error. Gyro drift depends on, among a few things, the inherent timing accuracy of the gyro time base. If this can be improved from the typical 50,000 ppm of the internal sensor clock to ~3 ppm using an external 32 kHz, 3 ppm accurate TCXO clock input then gyro drift will be commensurably reduced. It's just that simple...
Now 3 ppm might be a bit of overkill, in that even a 50 ppm crystal will be way more accurate than the internal LSM6DSV or ICM42688 clock...but 3ppm TCXOs are small, cheap, and ultra-low power, so why not?
