This is true, but my point is that shorter games don't have to be more snowbally. Heroes of the Storm, for example, has games that last 20 minutes on average, but also has tremendous comeback potential at all stages, in large part because it was designed from the get-go that way. By contrast, League's original average match duration was about 40 minutes, and merely accelerating all of that has caused champions to jump entire phases of the game ahead of others when they get a lead. Cutting down on the stat advantages these leads bring about would allow for closer games to happen, even if doing so would require a fairly complicated retuning of stats on items, champions and the environment.
You also hit the nail on the head in that, when snowballing gets this harsh, it tends to blot out skill expression in the game, and I think this also comes down to vertical power progression in matches: when a champion gets so many benefits from stats and levels that they essentially become as powerful as two or more of their opponents, they can mess up as much as they want and still get away with it, simply because of that buffer their stat lead gives them. By contrast, a player who's behind can play perfectly and still lose purely because they don't have the items or levels to match up to their opponents. It's the reason why every discussion on a champion's weaknesses or counterplay needs to include "... unless they're fed", and why classes innately prone to snowballing, like assassins, end up feeling horrible to play when behind, as they often cannot even succeed in the very situations they're designed to do well in.
It's because of this correlation between stat gains and snowballiness, and the inverse relationship between snowballiness and skill expression, that I'd be personally in favor of removing a lot of vertical power progression from the game, and only allow certain champions to gain more power as the match goes on: when a player gets a lead in a game like League, that should express itself by moving ever closer to destroying the Nexus via the destruction of structures. Otherwise, champions should remain on the same level of strength when fighting against each other, unless they're deliberately intended to gain or lose power over time. I feel one of the reasons why structures feel so weak now is because they're supposed to fall within a certain time window, yet most of a player's lead is expressed through their personal stat increases, rather than through the opportunity to damage turrets and the like, which causes them to need to be weaker, so that whichever lead goes into destroying them achieves it over a set duration.
When it comes to matchmaking, I feel that's a slightly different issue, in that currently Riot's matchmaking system goofs up a little too often and matches people of vastly disparate ranked divisions against each other. I don't think it's responsible for all the cases of extreme snowballing in League, but it definitely makes things worse by essentially guaranteeing that at least one lane will snowball as a result of that imbalance. This could be worth an entirely separate discussion, but I think one of the core issues with matchmaking right now is that teams need to be formed along a very specific, rigid framework (i.e. 1 top, 1 mid, 1 jungle, 1 bot support and 1 bot carry), and on top of that there are very large differences in popularity across those roles (support is still the least popular role in the game by far), which forces steep tradeoffs in time, accuracy and match quality. You could technically just autofill anyone to go anywhere at all times, but then people would hate autofill even more and would be forced into roles they're far less good at (which lowers the quality of matches), and making sure everyone is at the exact same elo/MMR likely risks incurring a steep cost in time, even if it is absolutely desirable when focusing on better match quality. This is one of the reasons why I think League needs to take a hammer to their years-old positional meta, as I feel it's directly responsible for a suboptimal matchmaking system, in addition to a large degree of stagnancy within matches themselves.