Sticking to monostatic radar, you'd be limited more by antenna directivity than the fact that they're smaller than the wavelength; weather radar operating at 2.8 GHz can easily detect insects up to about 10 km. Rayleigh scattering allows you to detect particles much smaller than lambda.
https://www.radartutorial.eu/01.basics/Rayleigh-%20versus%20Mie-Scattering.en.htmlHowever, to get a reasonable angular resolution (let's say, 10 degrees), one would need a ~1 meter diameter aperture - not exactly practical. Going to 60 GHz, you would only need a 5 cm aperture (at which point the approximations I used for beamwidth would likely fall apart). By sampling at a higher resolution than the beamwidth, one could fit a gaussian and get a better estimate of the target position in azimuth.
60 GHz parts are getting affordable, but phased arrays at those bands are still Bell-labs type stuff; I think they were able to demonstrate a TX array at 94 GHz a couple of years ago. You can buy a 28 GHz dev kit with 256 elements from Anokiwave/Ball Aerospace.
Finally, the really hard part would be figuring out how to reject multipath from the walls, roof, furniture, etc. Depending on mosquito density, living inside an anechoic chamber may not be too much of a compromise
