unstable IMU when powered with a battery

I saw something strange with the LSM9DS1 IMU on the nano 33 ble (works also with sense version), my circuit has a additional ST Mirco LPS33HW pressure sensor.
I cut the 3.3V jumper on the bottom of the board to power the circuit from something more "low power", magnetometer and accelerometer are calibrated and work fine on USB.

First I tried to power the howl thing directly from 2*AAA batteries (hey on the paper it should run from a coin cell). It worked really nice, but the LSM9DS1 hab become completely unstable.

Quite the same when the circuit is powered from a 1s lipo with 500mA buck converter. if the lipo is full it's kind of "OK" but once the battery is down to 3.7V (3.3V still OK) the IMU get's really noisy...

after some test's with other processors I came to two assumptions:

First: Powering a sensor over a GPIO is a really BAD IDEA.
The GPIO gives you between 1-4mA (assuming it's normal and not overdrive, 4.6.7 in nina-B3-series datasheet) and only the gyroscope is consuming on it's own 4.6mA of current in normal mode (3.2 in datasheet). I don't even mention the HTS221 and LPS22 powered from the same pin...
I think the additional resistance from the IO logic is enough to get the IMU to some unstable working conditions.

Second: Separated VDD and VDD_IO might could help. VDD_IO is coming from the "unregulated" 2*AAA pack and some buck or even LDO to power the sensors on a nice regulated 2.5V via VDD.
I will prototype some stuff next week and see if it works better

Has anyone experienced the same? the problem should even be reproductible with other ST IMU's like the LSM6DSM...

Back in the workshop I verified my ideas about the unstable IMU.
When powered over USB the VDD_ENV is about 2.2V when the IMU is active, far away from the 2.9-3.3V expected. Tension is just breaking down because of all the sensors powered from only one pin.
VDD_ENV goes down to 1.8V when the card is powered by battery (LSM9DSM1 works down to 1.9V). I tried several proven 3.3V buck and boost converters I used over the last years without success.

So this is clearly a "hardware-software" issue because the software is obviously using the pins with standard drive strength!

My solution is to tie the positive side of C13/C14 to +3.3V with some small wire.

Hope that helps other people how want to power this arduino with batteries...
Finally I use 2*AAA directly on +3.3V and now it works fine.