If a single interaction with the base chain costs more than setting up and running a full node for a year or two, then you don't really gain any decentralization benefit from keeping the current block size limit.
Excellent point! This would be a great metric to chart over time. tx vs full node opex.
Expanding block size would also increase full node opex while reducing tx cost (in theory). The wrinkle is that the lower tx cost may invite increased ordinals and inscriptions. The induced demand theory we see with highway expansion. Traffic only gets reduced temporarily as more cars fill up the space leading to the same or worse traffic congestion.