It's not like they're playing with the codecs. This has to do with resources, buffers, background functions or simple subtleties due to code and registry config. Ears are just very sensitive, especially if dealing with something that's inherently revealing and relatively neutral.
Do you think that iRiver/AK might have changed or tweaked the software digital filters in-between firmware versions? I grabbed an RWAK120-B last fall (2014) that had 1.40 pre-loaded on it, and it provided a very "liquidy" sound that, while hyper-detailed and nuanced, sounded slightly smeared/indistinct at the same time. I installed 1.42 about a month ago and felt that the sound became a little more lively, but it might have just been placebo as any difference between the two seemed insignificant as I listened to it more.
Anyway, I just got around to loading 1.32 earlier today (downgraded from 1.42), and the first thing that I noticed when listening to the same track was the gain/volume being less "hot", with more room to adjust the travel on my amp's volume pot. Overall, 1.32 seems to produce a more detailed and less weighty/liquidy sound, with faster transitions and improved dynamics.
Could it be possible that maybe AK configured 1.32 using a minimum phase software filter, which they changed (to a linear filter, or some other software adjustment) in newer firmware versions? Just curious if this might be the reason for the differences in sound from 1.32 to 1.4/1.42, or if there might be some other underlying reason for it.