There is more memory in the current lineup of speakers than there was in earlier models. And, for example, the Gen 2 version of the One has increased memory capacity over the Gen 1 version. The Gen 2 also introduced a faster processor. 

Changing it year to year would be challenging, though, it’s the ability to identify models which have specific RAM which allows easy identification as to what runs S2. Imagine if there were 15 different memory footprints in the PLAY:1, for instance. How would the average person be able to recognize which one they had, and which software version that each ran? Not to mention the need for Sonos to update all of the internal electronics to handle that change in each year, with new motherboards in each one. 

Isn’t this what S2 is all about?

Memory sizes have increased substantially over the years:

Right, Sonos would have to write code the lowest common denominator, memory wise.  There is little advantage to adding more memory to a Sonos One, for example, if you can’t write firmware to use the additional space.    Sonos will do this occasionally, which is partially why you see Gen 2 versions of products, but every year for each product means much more difficult management of firmware for each product.

I also wouldn’t be surprised if Sonos has multiyear contracts in place to purchase memory at certain market price, based on volume, etc.   Even if you can get more memory for cheaper, that you aren’t going to utilize, you still would need to go through some dev/testing to verify everything works correctly, adjust manufacturing, etc.

Partially, as I’m sure wireless hardware also played a significant part.  Probably other hardware that I’m not not familar with.