I'd like to see if someone can verify this: When mixing down 4K video tracks, the video loses resolution.
To replicate:
Create an UHD 30P project, populate it with an UHD clip, create a sequence, and put the clip on the sequence. Without making any other changes, do a video mixdown of that sequence.
When you compare the original clip to the mixdown, it should be visually identical (and on my system it is).
Now, change the project format from UHD 30p to 1080/30p. This adds a FlexFrame to the UHD material. Compare the mixdown to the original. On my system, the mixdown is much softer whereas the original is much sharper. It's easiest to see if you have a separate monitor, but it's quite visible in full screen mode as well.
For the mixdown, I'm using DNxHR HQ MXF. The source is XAVC-S at 100Mb/s.