Interesting thought which also begs the question whether streaming mixes can be targeted at cheap (in ear) headphones.

It would make sense since this is the way most people listen to music these days.

Knowing almost nothing about DSP is this something you can do?

For sure. The issue is the zillion variables in headphone mechanicals and transducers. Eventually someone will be making mems transducers in relatively cheap - acoustically inert - form factors and you'll just dial in your sound.

It's close with using hybrid anc but still too many variables to claim no sonic difference.

Definitely. Engineers historically would monitor in the car, in tiny speakers like Avantone Mixcubes, on even on boom boxes (I think Michael Brauer and Michael Beinhorn have gone on the record about doing it).

Today a lot of them are checking on Airpods and the like.

Definitely, and there are specialized plugins and EQ presets for this, as well for other media, such as vinyl. Have a look at this article about mastering for different streaming platforms https://www.izotope.com/en/learn/mastering-for-streaming-pla... (from a software manufacturer)

It might be the same mixing, but a different mastering.

On mastering it is completely usual to have extra compression/limiting, some light EQ and even extra "color". It is supposed to be subtle, but that hasn't been the case for quite a while.

Mastering is also when you get the final "bits" that go into a CD, so you need two different mastering sessions to get two different sounding CDs.

Fair enough. But consider this: If you're pressing 1000 units at a time, you probably don't have the budget to have separate mixes for each media. There is plenty of stuff edited by small labels (ex. transcending obscurity) on a tight budget.