> RAM size growth would probably continue.
True. But note - very long RAM grows ~ periodically doubling one chip size, and first chips don't have controller inside, so require very short traces to bus chip or CPU.
And usually, old chip becomes for example 10% cheaper, but twice size priced ~50% more than old, and to adopt new chips you need new memory controller with additional pins.
> At the time there wasn't really any established trend of progressive growth in graphics resolution or colors
Unfortunately, only partially true.
You may hear about RAMDAC on video forums topics.
It is partially palette, but also generator of video signal, reading from RAM very fast.
Problem is that first "fast page" DRAM have very slow interface, so when larger chips become available (and with cheaper kilobytes than older, this was real logic of semiconductor technology progress), speed of RAM was not grow.
And unfortunately, this once become bottleneck, it limits grow pixelrate, so even with twice RAM you could not got twice resolution.
In past, I few times calculated speed of RAM need to give classic 60 FPS, and at least up to (and including) first SDRAM machines just show their screen was enough to eat significant share of main RAM throughput, so internal graphics could even affect CPU performance.
On consoles problem was not so harmful, because limited resolution of consumer TV, but on few consoles used expensive frame buffer inside graphics chip.
On modern GPUs problem of RAM throughput solved by used overclocked designed VRAM chips and with extremely wide RAM bus, so chips run in parallel - in computers typical ~64bit, but GPUs start with 128 and top models have 512 or even 1024 bits.