Looking at the pinout for the NES and Famicom, it appears that the console exposes the chip-select and A10 the 2K RAM chip used for nametables, and then expects cartridges to feed those from whatever source would be appropriate. Would there be any difficulty with a cartridge strapping the chip-select low and then either wiring A10 to either PPU-A12 for games which want to split the RAM between the name table and tiles (which would allow four-way scrolling, albeit with "load seams") or exploit the RAM from unneeded screen rows for more tiles (use palette select to hide pairs of unused rows; each such pair would allow for four more tile shapes).
I would think this would have provided an easy way to minimize cartridge costs for games that didn't need too many tiles at once. The concept would seem like it would have been especially appealing in the days when SRAM was expensive, since games would be able to use CHRAM effects without any on-cartridge RAM. As a variation, a 4K ROM with an active-high chip select (used on Atari 2600 carts) could have been used to allow games to have 128 ROM sprite tiles, 128 ROM nametable tiles, *and* 64 RAM tiles for either usage at no extra cost. Have any games ever done this?
I would think this would have provided an easy way to minimize cartridge costs for games that didn't need too many tiles at once. The concept would seem like it would have been especially appealing in the days when SRAM was expensive, since games would be able to use CHRAM effects without any on-cartridge RAM. As a variation, a 4K ROM with an active-high chip select (used on Atari 2600 carts) could have been used to allow games to have 128 ROM sprite tiles, 128 ROM nametable tiles, *and* 64 RAM tiles for either usage at no extra cost. Have any games ever done this?