The answer is application dependant, alas.
The best you can do is to design it well, take account of possible issues in advance, and then track down and fix the 0 or more ones you missed.
Taken at face value this is eminently doable.
A few things that come to mind. There will be others.
4 stroke engine ignition will radiate some noise and you may also get some conducted noise. This can be checked for early on. An AM band radio held in close proximity to potential EM radiators and tuned across the broadcast band makes a surprisingly good EMI detector. An FM band radio is nowhere near as good for this.
The alternator should be relatively clean. The waveforms rend to be sinusoidal only at low speed and become trapezoidal quite early on under load as the core material starts to saturate. You might get diode switching noise - this can be quite bad when it happens, although this is less liable to be an issue with 3 phase waveforms where each diode "hands over" to the next in turn (as opposed to single phase where a rising waveform suddenly starts dumping into a capacitor that has no other supply source at transition time).
The battery acts as a gross filter but cannot be relied on to take out all the "more interesting" noise. If necessary an inductor and capacitor L filter in the ESC feed will help. The ESC will have onboard power supply bypassing. Adding more MY be useful.
The ESC either uses sensors or (more usually nowadays) uses sensorless speed detection using winding back-EMF detection. Noise on the sensor inputs is a potential problem - not one I'd overly expect, but worth being aware of the potential (pun noted) if things don't work.
