You could refer to this post regarding the impact of using an external crystal with the BNO055.
BMF055, BNO055 and BNO080 all share the same hardware. The difference is that BNO055 comes with BSX3.0 sensor fusion software from Bosch Sensortec already built-in its embedded microcontroller while the BMF055 does not include any software, thus its Cortex M0+ microcontroller is freely programmable.
The BNO080/085 (and previously the BNO070) are parts programmed, sold, and supported by Hillcrest Labs. Please refer to Hillcrest directly for support about these products.
Many thanks for the answer.
"The choice of using a crystal or not in the microcontroller is up to you, and the benefits of a more accurate clock vary per application."
This is about that what Adafruit writes. The accuracy is probably most interesting when it comes to "drift reduction", so if you have some rough percentage estimates of drift differences (with/without crystal) at more or less fixed temperatures, it would be great to hear. Herbert (https://community.bosch-sensortec.com/t5/MEMS-sensors-forum/BNO055/m-p/7730/highlight/true#M638) was investigating drifts, but unfortunately in this reply he didn't compare the case with and without crystal.