A lot of the time it does affect the quality of a game though, at least for some players like me. It affects the overall feel of play when you see reels spinning with symbols that never actually land at anything close to the frequency it looks like they should, not to mention if they do stuff like landing scatters just above or under. It is perfectly possible to make a complex game while still giving a good visual representation of actual chances of hitting the different wins.
Of course there are lots of game developers taking shortcuts and in the end producing bad games. One of Netents few decent games in past years is Jungle Spirit, and it still suffers from the awful random feature, has anyone ever won anything on this past 30x bet? If you're going to give the player a 10-30x win, then do that, just throw the coins out, don't fake it and pretend the reels spinning matter. One of the worst on abusing this is Elk, pretty much every single game of theirs feels like it just decides on a win and then finds a spin that matches it. The same game could be done with just doing better maths and letting the reels decide and it would just feel better to play, but I guess the amount of players that think that way is small enough to be ignored?
There are some good examples of games being done right out there, for example 3dice have some where you clearly see the odds in features with a ball dropping in a series of 50/50 chances, like Medieval Moolah and Fortune Falls. I'd be very surprised if those games use any kind of weighting for the reels too.
BTG games just feel right, you can look at the reel strips that have been datamined and compare to your own results and you can see that if the stars align you can win really big, not being capped by some arbitrary number some game developer thought was good enough. If you need something to be rare, just make 150 stop long reel strips like Donuts?