Hi @ufocet,
This thread is likely worth a read.
One of the noteworthy points is that the back-EMF induced from braking in the air is generally more significant than in the water, so it’s possible your bench test setup is running into issues that would be less of a problem for an actual vehicle in the water. That said, computing devices tend to strongly dislike large power fluctuations, and the 1.52 V range shown on your oscilloscope does suggest you likely want either a more stable voltage regulator, or some kind of filtering circuitry added to your current one.