Exactly, you "turn the JPEG quality down a touch". That's a problem for two reasons. One, you're reducing the file quality (which may not create visible problems initially, but - I promise you - will cause all sorts of ugliness to appear if you post-process the frame several times over, as is par for the course in professional film and video production). Also, you're shooting JPEG, not RAW, meaning you're comparing apples to oranges. Try that test again using a file format comparable to what D-Cinema uses. Again, 3-5 frames is on the high end, and WAY short of the sustained 24 fps needed to compete with film.
Keep in mind that a 400' film load (a standard measure) gives you 11 minutes of shooting at 24fps. Every time you stop the camera to swap magazines, you cause a break in the shooting. Producers want to decrease the number of interruptions like this, not increase them. One of the (many) things video needs to offer to be competitive is equal or greater time between reloads. With digital, memory capacity isn't the only limit on run time. The reality of cameras shutting down due to overheating (a non-issue with film, but something I've experienced with Canon 5Ds) also needs to be accounted for. And again, this is all about beating established industry benchmarks. On a high-end shoot, your camera package is actually one of your smaller costs. Saving a few bucks here is bad economics when it leads to delays that put your crew into overtime.
Regarding "Toy Story". That's a CG film. At the time it was made, per-frame rendering times for (a) the source images (b) the up-sampling and (c) the film-out were all measured in minutes, hours, and sometimes days - not fractions of seconds. Because the production wasn't having to contend with anything happening in real time, they could afford to take the extra time to create those initial frames using ridiculously high bit rates and no compression whatsoever. In other words, they were free from the two primary sources of trouble that lead to visual artifacts in upresed images. And since the entire film was created in-computer, there was no post-processing whatsoever. That is to say, the upsampled files were first generation copies of (exceedingly clean) master images.
It's worth remember in that the massive render farms used to deliver all this are exactly that - massive. As in, they need their own air-conditioned warehouses. That kind of hardware support isn't even feasible for studio shooting, let alone location work. So sure, you COULD master at 1k and go up, but only under incredibly controlled conditions that result in bills averaging $1 million per finished minute.
Shooting action live then heavily processing your material in post is a completly different ballgame. Personally, I'm thrilled that video solutions are coming to market that can (finally) rival film, which is a colossal pain in the ass, and the source of some pretty major creative constraints. Still, the competition had some cons that were (a) major and (b) non-obvious to non-professionals. So I wanted to help mey understand why the change from film to digital hadn't occurred with the same speed and thoroughness as it did in the still-photography world.