I think there is a lot of over-estimation in here on how important memory bandwidth will be.
An RTX 3050 4GB mobile has about 70% more (1.7 times the) memory bandwidth of an RTX 2050 4GB. Otherwise they are identical chips. What is the difference in their relative performance at the same resolution (1080p) and TGP (40W)? At most 12%. On average 4.2%. Is that a meaningful bottleneck? I'd say no, but you might say yes.
https://www.youtube.com/watch?v=MIimIAMv7sk
These are the same architecture as the T239, so it is as roughly a like to like comparison as we can get.
At the 800p/900p/1080p native resolutions that the Switch 2 likely will target in docked mode, I don't think memory bandwidth is going to be the main bottleneck. Memory capacity (like with the Series S and these low-end Nvidia laptop GPU's) is probably going to be a bigger bottleneck as we've been seeing with recent titles (Alan Wake 2, TLOU Part 1, Hogwarts Legacy, etc.)
Still I think Nintendo has gone a pretty balanced route, assuming these specs are true. Nothing seems extremely lopsided. I expect most Switch games to target native 800p/900p in docked mode and upscale to 1080p or higher effective resolutions using DLSS, given these details. That was more than I was expecting a few months ago (mostly sub-720p -> 1080p in demanding titles.)
Should put it about a quarter-tier above the Rog Ally and a quarter-tier below RDNA 3.5 (Strix Point) handhelds, which probably will release late this year/early next year. I am expecting those RDNA 3.5 handhelds to be competitive with the Series S in terms of performance when at their 30W-40W power profiles and in devices with good thermals.
