OK. Quite a bit of experiments follow:
Power, current, CCx lines
Using a PD analyzer which measures VBUS voltage and current, as well as voltage on USB 2 data lines and PD CC lines, I captured the first few seconds of connecting an iPad, a Macbook and the Framework 12th Gen to the LG UltraFine 5K using Apple’s TB4 Pro 2m cable. Both iPad and the Macbook connected and displayed content on the monitor without issues, every time. Framework failed every time.
Macbook
CCx lines
D+/- lines
iPad
CCx lines
D+/- lines
Framework 12th Gen
Overview
Framework keeps going through a loop. This is an overview of the Voltage+Current. Further down, only the part in the middle where there is a dip is zoomed in and shown.
CCx lines at the dip
D+/- lines at the dip
USB-PD packets
Right before the dip, Framework asks the monitor to enter alt mode and right after it receives the acknowledgement, there is a detach and attach event which seems to be equivalent to pulling and reinserting the plug.
In contrast, Macbook continues on without issues:
Since this happens even when the system is turned off, it must be a bug in the EC firmware (or in the hardware). I spent a short time going through the firmware code guessing where this may be happening. I think it may be right after the USB mux is configured for Thunderbolt that the detach happens.
EDIT: hmm, maybe not, it takes more than a second for the detach to happen since the alt mode request.