That's interesting. With a sampling rate of 88.2kHz, the frequency is 32766Hz. With a sampling rate if 44.1kHz, that is above Nyquist, so it gets aliased down to 11334Hz! When using the audio inputs, the sigma-delta converter filters it out appropriately, so you do not hear it (see the first link I posted above to see more on this).
No idea what is the source of the interference though ... do you have some ultrasound emitter in the room where you are? Or maybe your laptop? It could well have an electrical origin, though. Could you try plugging the 3v3 line into one of the inputs and see if that frequency is still there? A high-frequency noise on the 3v3 line combined with the biasing circuit of your preamp (which simply divides it by two to generate a virtual ground) could well cause this problem. If that's the case, try using the (unpopulated) MICBIAS pin on the Bela cape instead of the 3v3 line as a power supply for your mic. I am not sure it will be able to power all of them at once, but it should be enough for testing with just one.
Anyhow, for practical purposes, you could probably just filter out the offending component (with a notch or low-pass filter).