I doubt there's a problem with the EEPROM in those motes going haywire, or something being wrong with those atmegas. I have used SMs since inception and they still work flawlessly.
It sounds like a button is stuck or intermittent somewhere.
That causes the SM to sync and ultimately delete the sycn info. I have seen this when my kids played with my SM's and kept them pressed.
That could explain why you're seeing no data (empty?) EEPROM and them being out of sync.
That is controlled by this setting:
#define ERASE_HOLD 6000 //time required to hold a button before SYNC data is erased
I don't think the gateway itself will help a lot in debugging. SYNC'ing is completely independend and only requires the SM's themselves. When they SYNC they will only talk to each other, and then of course inform the gateway when they do turn ON/OFF after SYNC'ing. Otherwise the gateway will not "spy" on that activity or report anything other than direct messaging between itself and the SMs.
So not sure what to suggest but here's something:
Try to take the caps off those switchmotes. It's also possible the lasercut holes in the acrylic (+the wear of the buttons?) causing the buttons to become easily self-pressed. Just theory here but I thought I've also seen this happen a few times but could not repeat it unless I tried really hard (and my SM's are far older). So take the caps off and try to sync them again. Maybe first erase everything with the 6 second hold to start from scratch.
FWIW - I need to simplify how they sync and how the SYNC mode works, but just haven't gotten the time to do it. Namely I now feel the SYNC mode should by default fully sync (ON and OFF) two SM's with a long-press on each button. But this is kind of beside the point of this thread...