I think your analysis matches my experience using the 12900KF working with large projects in Cubase. I noticed that the lower/faster settings on the buffer quickly need adjusting as I add more Virtual Instruments - but there seems to be no limit to how many I can add as I increase the buffer. A project I remixed that would crash my old i7 8600 now runs smoothly and only uses 24% of my CPU, and I added many complex plugins, like East West convolution reverb. It is a dream machine for virtual instruments, which is perfect for me since I only play the bed tracks in simple piano and then get to work from there.
I believe the new gen of intel and any future hybrid cpus pay the price of having different cores and an AI to interpret the work load with some additional latency. The 5950x and the previous gen of intel CPUs the 11900 are the last of the simple high performance low latency chips. These new gens with increasingly hybrid designs and chiplets all pay a price with latency but you gain big workload returns for huge projects. I was able to unfreeze all 120 tracks without any issue. It’s just astonishing.