The younger audience of today is probably having hard time comprehending the concept of sprite that pretty much defined real-time imagery for more than a decade or so between the 80s and 90s. Sure, we do still use the term to denote planar 2D images representing 3D objects in a 3D space, but it has a very different meaning today than back in the day.
In the era of 8 and 16 bit systems, like the Commodore 64, Super Nintendo or Neo Geo, sprite denoted specifically an independent, dynamic visual object (like a space ship or a race car) of which properties, such as size and number, related directly to the capabilities of a certain piece of hardware. So, for instance, the C64 could display no more than 8 sprites at a time which paled in comparison to the 96 sprites the more advanced Neo Geo was able to push to the screen at once. Put simply, sprites were Lego-like building blocks for the dynamic elements of the game, and very much of the visual outcome depended on them, whether one liked it or not.
As said, not only the number of simultaneous sprites varied between the systems, but so did the sizes of them as well. That is, how many pixels one sprite could consist of in maximum, which was relatively few in lower-end systems, such as the C64. This limitation made larger game characters and other dynamic objects that more fascinating at the time, even if the assets usually consisted of several sub-sprites to make an impression of a one large one.
An iconic example in my mind is The Way of the Exploding Fist (1985) originally developed on the C64 that was, in addition to the great gameplay, celebrated indeed for its seemingly massive characters, which, of course, consisted of a number of separate sprites. Nevertheless, the size of the characters in and of themselves bore so much aesthetic value that the technical side was secondary: big was big was big.
Since the fascination of real-time imagery is very much based on the recognition of technical limitations and, at the same time, pushing that envelope of what is considered possible, it’s not coincidence that some games, especially the ones released on the launch of a new hardware platform, exploit this frame of thinking. The thing is, there is this short post-launch window within which a game can make an impression employing merely the basic features of the freshly released platform.
One instance that comes to mind is Super Mario World (1990) that came bundled with the Super Nintendo as a launch title. So, I believe the sole reason why there was a huge Bullet Bill, aka Banzai Bill, right at the first level was indeed to impress new console owners, or people playing at stores, with the power of the SNES. And since the sizes of game characters were highly limited in the previous generation of hardware, it was only natural to put an oversized version of an already established character to mediate the point across: Kids, we can have this big Bullet Bills from now on.
It was the inevitable decline of sprite-based hardware that made the issue of size obsolete in the realm of real-time imagery. Once everything was constructed with polygons, the dimensions of art assets became totally relative, and thus a non-issue technically speaking. Which seems to escape people who marvel, say, God of War III (2010) as a technological achievement for its colossal characters. I can too scale up a 3D model in a 3D software environment as much as I please and technically it makes no difference whatsoever.
If, then, one “superficial” attribute has to be singled out that bears any technical meaning in art assets today, it’s definition instead of size.