@Ephemeral: Is it possible to check the playcounts of easy diffs on songs with different lengths? Basically if there's a significant difference between the amount of plays a 2 minute easy diff and a 4 minute one would receive or not.
I think it would be pretty useful information in regards to Shadow1and's suggestion (I actually came here initially to post the same idea). The reason people want to get over the approval length limit and "abuse" it right now, is because the amount of work required is significantly reduced just from a few seconds of difference. A 5 minute map requires you to map less total time than a generic 5 diff spread for a 1:30 tv size, and while obviously this doesn't translate to the same amount of effort needed (mapping lower difficulties are tend to be faster because they require less notes to be placed, and are more restrictive anyway), I think it's clear why people want to get to the approval length when they're really close. Arguing about laziness and whatnot is irrelevant and should not have anything to do with a ranking criteria anyway, and even then it's not true for a lot of mappers. For instance, I'm also a player, so naturally I split my time between mapping and playing. Adding to that any real life obligations I have and whatnot, I might just not have enough time or will to map and maintain a full spread for a 4:58min map in hopes of it getting ranked. My efforts would be much better spent on either extending said song, map something else, or just play the game instead. Also, if we're speaking about laziness, then the 8 diff rule doesn't exactly promote unlazy behaviour anyway.
And I think the reason is just that the difference it makes is too big, so the rewards for "abusing" it is higher than just mapping the spread. Instead, making it gradual would make the whole thing less of a pain in the ass for mappers. For example, maps above 4 minutes need 3 difficulties, 5min need 2 and 6min need 1 (I'm just pulling random numbers right now). If we can look at the statistics of the playcounts of different difficulties and their map lengths, we could maybe find reasonable lengths to make these cut-offs, and which difficulties could be excluded. This way, the gain from extending an .mp3 is less compelling than before, and also encourages mapping for a wider audience in general.
Although I personally don't find extending .mp3's a problem in the first place. In a lot of cases you probably wouldn't even notice it was unless you've heard the song before, or someone tells you.