Ah, my bad.
For one I left LOG=core enabled, which can be a problem on some systems, and also the image.elf.gz could be too large to boot now. We already encountered this problem on other machines (amongst others an AMD ThreadRipper based on).
So, I’ve made two new images:
❶ image (sig) with the NVMe and usb_net changes w/o LOG=core
❷ image (sig) like ❶ but without any pc_wifi firmware, i.e., the image.elf.gz is smaller but wifi will not work w/o firmware.
If both images do not work → back to the drawing board but that will have to wait until after FOSDEM (FYI, I have not checked if the images boot on hardware, as I am on the train right now, but they should).