Why are you convinced it was a step backwards?
Unlike HFS+, APFS was designed around SSD performance characteristics. Therefore, its on-disk layout assumes that seeks are more or less free, that there is no decrease in seek performance with increasing seek distance, and that data placement mostly does not affect throughput. None of these three things is true on HDDs. HFS+ on-disk layout, on the other hand, is well-tuned for these HDD quirks. Fusion Drive's original implementation was also tuned to work well with HFS+ (or perhaps it was the other way around, Apple could have put in hooks to let the HFS+ filesystem driver guide data placement).
It's very plausible that when Apple began experimenting with running APFS on a Fusion drive, they found that the old HFS-tuned tiered storage mode didn't work well, and that a more cache-like approach performed better. The point of Fusion was always making a large HDD perform like a SSD, especially when your frequently-used data fit entirely on the SSD. Sacrificing a little capacity to gain performance fits with the concept.
The RAM tier here is redundant as the macOS virtual filesystem layer already caches file data in RAM for all filesystem types on all media types. (Also note that you really don't want RAM to be treated as a tier rather than a cache. What happens if the filesystem is very full, meaning there's data in RAM that there's no space for on permanent storage, and the computer has to shut down?)
There also isn't much point in adding a tier that's too close in performance to the tier below it. This is because you don't want to pay data migration costs without an adequate performance gain. Putting an extra SSD layer in there seems like exactly that problem. External SSDs perform close enough to internal that it's probably not super worthwhile.
