What dithering mode for photos, and how much should I pre-process?

Both camps are half right. What matters is that someone makes the tonal decisions deliberately, and the software's defaults don't know your material.

My workflow:

1. Crop and straighten, convert to greyscale.
2. Compress the tonal range to what your material can reproduce. Wood does maybe 4–5 usable tones, anodised aluminium 3, slate a generous 6. Lift the blacks, pull the whites down. It looks flat on screen, correct on the material.
3. Boost local contrast slightly (unsharp mask, large radius, low amount).
4. Then let the machine dither. Jarvis for general work, Stucki for faces, ordered/Bayer only when you deliberately want a visible pattern.

Pre-dithering to 1-bit yourself is only worth it when you want a specific halftone look, because then you control dot size and angle. For normal photo work, doing tonal prep well and letting the machine dither gets you 95% there with a tenth of the effort.

"Compress to what the material can reproduce" is the thing I was missing. My exports were full-range and everything below 30% grey just came out as solid burn.

Also test at final size. A face that looks great at 40mm turns to mush at 15mm because the dither dots stop resolving.