"They just work" is a surprisingly important start. FreeCAD improved a lot with it's latest release, but it still has many flaky gaps. Caps / apices are especially hard.
Higher order continuity native to the blend would be a requirement for many, too, and the algorithms only get harder and trickier to verify correct.
I intend to open source it once fillets work real good. That will take a while, but I've made pretty good progress since I started 10 weeks ago.
Medium term goal is to release a GUI application that can be used to import STEP files and just do very high quality fillets in. Would be very useful for many hobbyists, I think.
I just bought my ppu digitizer two days ago! Very excited to get that working.
There are a couple reasons you would go this route or classic nesrgb, and I was very tempted (I might still do a nesrgb on my other nes, if ppu digitizer goes well).
Primary one is lumacode is a digital signal that requires interpretation before it can be input into a tv. You can use the creator's rgb2hdmi which requires a raspberry Pi, or you can use a scaler like retrotink or ossc (pro). An RGB connection is standalone in many setups, and also works with scalers. If you want to interface with a CRT, lumacode doesn't immediately support that.
Additionally, ppudigitizer doesn't (directly) support multiple palletes. You are stuck with whatever your lumacode interpreter provides (though you could in theory change this yourself).
Higher order continuity native to the blend would be a requirement for many, too, and the algorithms only get harder and trickier to verify correct.