I guess, you are asking RF engineers as a biologist? This way you should present an easy to understand drafting of what you need. And probably cooperate with an active RF researcher who have both, good simulation system and RF power test facility in close access. You should look for such in your university first. Now consider these factors:

-You remember those small holes in the window mesh. That mesh is a Faraday cage. Holes should be much less than a half-wavelength. The wavelength is somewhere around 12cm depending on humidity and gas composition. So normally we make it 1/20 for low-power quick tests, and 1/50 for good screening. But for complete safety manufacturers make it 120 times shorter. Why? The field kind of sticks out of every hole even if not radiating further.

-the door is electrically connected to the chamber with carbonated resins, or sometimes with spring contacts. If contact is lost, there is also a tricky wave-path design in the door sealing (flange width, and sometimes a wave maze, analog of 1/4 slots/grooves), which makes the leaking wave cancel (reflect) back into the chamber.

Don't break both of these.

Do not make big changes into the chamber itself, because you can change impedance matching with magnetron, and burn it down.

Now, considering small apertures, you can probably use them up to 1/8-1/6 wavelength size. If there is a good conductive isolation of the pathway and a safe no-pass environment. You could probably add a corrugated notch waveguide filter at the input, but I think it is impractical with such large wavelength.

Coming back to engineers. An experienced guy can find you a spot with a weakest field in most heating conditions in the chamber, good for a wide material feeding window. But both simulation, measurements and trials (in safe environment) are needed.

Technically, it can get stuck into full microwave oven reverse engineering/design worth of hundreds of thousands of dollars.

Thank you. I completely agree with you. I have to involve RF engineers.