The reason why bandwidth shouldn't be a bottleneck is that internal resolution isn't really a choice of discrete values that need to scale well to output anymore. You can arbitrarily set the internal resolution or use (a good implementation of) DRS to have it scale to the appropriate value where memory bandwidth is no longer a bottleneck.
The memory bandwidth of Switch 2 docked and handheld is roughly in line with the measured optimum of about 33 GBps/TFLOP for Ampere chips. So compute capabilities and memory bandwidth likely are well-aligned on the T239 if the developer actually scales workloads appropriately. This isn't like a Rog Ally situation where the paper compute (excessive of the Series S) is heavily limited by not having enough bandwidth.
Likewise, if Nintendo fixes their VRR, framerates shouldn't necessarily be discrete either and it is probably that reason why Switch 2 could have higher unlocked framerates than a console that doesn't support VRR at all (i.e PS4.) If Nintendo fixes VRR, it might not make sense to lock to 30fps in many instances because the people who care a lot about performance likely have VRR-capable displays, and the people who don't might not notice screen-tearing and other issues. So long as there aren't very sharp drops, of course.
