I have not encountered any of the issues you report, using 2048 for CFF OT fonts. The only issues I encountered were a) half height cursor selection colouring in older versions of InDesign (a cosmetic problem), and b) problems with scaling in some third-party PDF creators (a serious bug, and the reason I regretfully went back to 1000 UPM for CFF fonts after that).

I had some @font-face issues with a 3k UPM.

and the reason I regretfully went back to 1000 UPM for CFF fonts after that

Why is it always font designers who have to work around bad software? Wouldn't it make more sense to just start releasing valid CFF OpenType fonts and tell software developers to get their act together and fix their crappy code?

James, that's usually my preferred approach: build fonts to spec and log bugs against software that fails. But in this case my clients wanted to avoid the PDF creation bug, because its so bad.

I've not done the unthinkable then by upping to 2000?
Interestingly the halving in InDesign is only suggested in the panels, not the layout, so a kerning value of 100 in FontLab will show as 50 in InDesign type panel but 100 in the actual layout…

I don't want to go back to 1000 as it doesn't suit the way I like to draw.

the halving in InDesign is only suggested in the panels, not the layout, so a kerning value of 100 in FontLab will show as 50 in InDesign type panel but 100 in the actual layout

Yes. Adobe by default apps express all kerning and tracking in 1/1000 units per em, regardless of the actual UPM of the font in use. This is sensible, because it means that spacing is not radically scaled when you switch fonts.

What I do know, Truetype can working in 2000 UPM around and OpenType usually at 1000 UPM. If OT upgraded to 2000 UPM, you will find the problem.

Best