Why are the SCL and SDA pins by default (nRF SDK 11.0, nrf_drv_twi.c: 44-57) configured to use the internal pull-ups?
At 100 KHz internal pull ups will be fine. At 400 KHz the speed may become slower (depends on the total capacitance and resistance of the line) if only the internal pull up on the nRF is used, this is at least what we have seen if you connect the one nRF as TWI master and one nRF as TWI slave and only have internal pull up on one of the pins (the clock line became around 250 Khz or so). If internal pull up is enabled on both the master and slave, 400 KHz should work just fine. I will update my answer to be more precise.
Internal pull up something totally separate than the TWI peripheral. You can see the internal pull ups as a "hack" you can use to remove the external pull ups (and hence save board space and BOM).
At 100 KHz internal pull ups will be fine. At 400 KHz the speed may become slower (depends on the total capacitance and resistance of the line) if only the internal pull up on the nRF is used, this is at least what we have seen if you connect the one nRF as TWI master and one nRF as TWI slave and only have internal pull up on one of the pins (the clock line became around 250 Khz or so). If internal pull up is enabled on both the master and slave, 400 KHz should work just fine. I will update my answer to be more precise.
Internal pull up something totally separate than the TWI peripheral. You can see the internal pull ups as a "hack" you can use to remove the external pull ups (and hence save board space and BOM).