Scanning pictures, either from enlargements or film is AD-conversion. In most other AD-conversion cases people know that they need at lest double the digital sampling rate to get a decent recording of the analog signal, at least double. But very few view scanning as the AD-conversion it actually is.
Take for example a good film that can capture 60 lpmm, that is 120 pixels per millimeter. Now take twise that resolution to get the double sampling rate, that is 240. 240 pixels per millimeter is 6096 ppi! And this was only the usual low contrast measurment, at high contrast like 1000:1 films can have much higher resolution. If we scan 24x35mm in 6096 ppi we get an image that is 5760x8640 pixels, that is almost 50 megapixels! Not that a small frame image contains that much, but to get it without strange aliasing effects that can exaggerate grain we need to scan at that insane(?) resolution and having at least 48-bit color. A 6x6 medium format would be 207 Mp scanned at this resolution, that is a file size most computers would have some trouble with.
Now lets take a look at scanning an enlargement, 30x20 cm. What resolution would you need to get a 50 Mp image? 800ppi will be more than enough. Can RA-4 paper hold 800ppi? Yes, barely, but it can, the paper structure will affect it to some grater or lesser degree. So, if you do not have access to a top notch drum scanner capable of 6000 ppi or more, you will probably be much better off making an optical enlargement on a good paper, then scanning the large copy. Yes the enlarger optics will degrade some, the paper will also degrade some but also smooth grain in a very natural way, the net effect will in most cases be better than a 3000 ppi scan, not so much regarding detail, rather having about the same detail but with less and more natural grain.
I learned this the hard way when i was beffled by a less grainy Fuji superia iso-400 scan at 3200 ppi than an iso-100 Fuji Reala shot at the same occasion with the same camera. First i could not figure it out. When i looked at the optical prints with a loupe the 400-image was a lot more grainy, but scanned it was the other way around. This is the Nyquist frequency playing tricks on us, a reality known for a very long time in digital recording of sound, but the exact same principles is at work when scanning images.
Think of it this way, just to wrap your head around it. Imagine 3 "grains" in a row, black, white and black.. Now scan this with 2/3 of the grain resolution. The first pixel will see one black and a half white, averaging this to light gray. The next pixel will see a half white and a whole black, again averaging this to the exact same light gray. Now imagine this to be horizontally, but the same to be true vertically.. congratulation, now you have big blob of light gray instead of 9 pixels in a checkered pattern, shock horror, a huge grain.. at least it looks like that.
Now this was only BW, the same goes on with colors, that is why we need a huge bit-depth as in 6553600 colors per channel. But most scanners can do that, but not many image editing software and certainly not our monitors. That is why serous scanner software needs a special definition for every combination of scanner and make of film there is, especially when talking about negative film that has a lot of exposure latitude, and of course olso the orange mask. But very few scanning software have the ability to be configured with different medium definitions.
When people are comparing scans, they are most of the time just comparing the scanner technology, not the scanned original. When people said that digital passed film at 4 Mp, it was only the consumer scanners that could not get more out of the film, later it became 6 Mp and then 8 and 10. The film was the same, only scanners became better.
