But at the same time, we're sacrificing all these cars for.. What? The ability to 'open them up' and view all the metal covers and wiring and such..?
I'm with you on not pushing graphical prowess over good racing and the like, but I sort of doubt how much the two overlap in a development sense. There's always the budgetary side, in that maybe they spent all their money on graphic artists and then got an intern to knock up the career over a weekend, but for the most part I feel like graphics and gameplay are fairly independent.
That may not be the case, but that's sort of the assumption behind what I say.
In terms of why you're sacrificing all the cars...I feel it's better to get it done in small chunks. If you look at GT6 now, it's got 1200 cars and presumably about 400 of them will be premium. Say they get another hundred or so out as DLC between now and GT7, and they have another hundred or so held back as new content for GT7.
They have to drop from 1300+ to ~600. That's a big drop, even if they go the sensible route and start consolidating their dupes and near-dupes. And that's only if they're still using PS3 era assets on PS4, so while they look very nice, technically they're still behind the times. Who knows whether they're built to a level that enables the kind of internal detail (and potentially damage) that it looks like Forzavista models will be capable of.
Forza on the other hand likely keeps maxing out at 6-700 cars at the end of each generation, but those cars are built specifically for the platform. How much of an advantage that is is debatable, but it seems like a better idea to me.
Personally, I don't have as much of a problem with lack of raw car count as some. I was perfectly happy with the car list in say Shift 2 (car list only, there was other major issues with that game). ~200 well chosen cars is enough for me to have a good time.
More is always better, but I'm not really that willing to accept lower levels of quality for that more-ness. 200 premiums in GT5 (had they been well chosen to cover a wide range of types) would have been acceptable, had it taken 2-3 years to make instead of 5.