The rPi foundation doesn't care about the individual end users anymore. They make most of their money supporting commercial vendors. They're keeping up the appearance of being good to the little guy so they can keep their support ecosystem alive and well.
(I say this using a cm4 for image acquisition commercially. Because of the way our pipeline works, I skip the hardware encoder anyway - first, it's a complete pain in the ass to use, and second, it can't handle full-resolution images from a pi camera. We're slower than video because of the resolution, but still capture between 1fps and 1/10fps depending on configuration, with enough headroom to do some on-device image classification using tflite, rotation and JPEG compression, etc.)
Obviously, it would need benchmarks, but taking that post at its word, the 5 can transcode twice as fast in software as the 4b could do in hardware.
That doesn't mean some other hardware platform wouldn't be drastically better, I just don't get the extreme reaction about the 5 when compared to the 4.
Do you have evidence to the contrary?
[0] - https://www.raspberrypi.com/success-stories/korg-synthesizer...
https://www.raspberrypi.com/app/uploads/2023/04/230329-Yodec...
Yoto Player is another example but I don't know how many Pi-based ones they shipped, I'd guess more than 10k but I don't know.