I've kind of gotten at this a couple times previously in the thread but I think the snowballing discussion is kinda missing the forest for the trees (the stated aim of the changes attempting to curb snowballing made in 7 wasn't to address any mechanical aspect of game balance arising from snowballing itself, but to mitigate the late game being boring by reducing the likelihood a win is assured by like 1400 BCE; snowballing nerfs were the means, not the end).
Regardless, I do think it's interesting as a topic. Snowballing is kind of inherent to literally any game played over a fixed duration where you obtain things, since the sooner you get an item the greater the percentage of the duration it can benefit you. I don't think it's inherently good or bad - it's just a question of what impacts it has and how they affect the player experience.
To give a few examples of other games where it exists in one form or another:
Slay the Spire is probably the most "snowball-y" game I've played. It's a roguelike where you assemble a deck of cards (and some bonus items called relics) over the course of the run, from which you draw attacks to use in combat encounters. "Good now" is massively better than "good later", since the sooner you get something, the better your chances at opting into harder encounters, which give better rewards, creating a compounding effect. The few things in the game that have delayed effects are generally regarded as being rather weak since a delayed dividend is so unfavourable compared to something that starts benefitting you from the very next encounter. I don't see snowballing as a problem in Slay the Spire because it's a big part of the power fantasy of roguelikes. Lean into that feedback loop of strength begetting strength is not only fun, but outright required to beat the endgame bosses.
That said, the risk of making late-game choices meaningless is still present, and I think the devs get around it in three main ways. Firstly, the game's difficulty assumes you're snowballing - since you fight different enemies over the course of the game (unlike civ), they have the luxury of throwing harder stuff at you, so your win (most of the time - you can outscale the game on a very lucky run) isn't assured until the credits roll. It'd be like if you had a fantastic exploration age and then all the AI leaders despawned and were replaced with "Mecha-Caesar" and "Ascended Lincoln". Secondly, the vast majority of loot in the game leans into the snowballing. As mentioned above, there are only a handful of "good later" items that exist. Most of the time, the choices you're faced with are one "good now" vs. another "good now". Third, and most importantly, Slay the Spire is built around items synergising with each other. A winning deck is typically greater than the sum of is parts, and what that means is that the addition of any single card can be massively impactful. Choices late in the game still matter because you could see a card that completely transforms your deck.
For a worse example, Dead Cells is another roguelike (or roguelite; some people are particular about that distinction) aiming for the same general flow - pick up loot to become stronger over the course of a run. The problem is that Dead Cells has a massive loot pool and every item is sorted into one of three categories, each corresponding to a stat on your character. These stats scale with scroll items you find as you explore, but each scroll you find only lets you choose one stat to raise, and the scaling works like compound interest. What this means is that the minute you use your first scroll, you're basically locked into the stat you chose to raise, because spending even a single scroll on a different stat past that point will result in missing out on hundreds of percentage damage by the end of the run, against enemies who are tuned expecting you to have focused a single stat. This means that two thirds of items - everything that corresponds to a stat you didn't pick - become actively detrimental pickups from the minute you find your first scroll in the first biome. There's no potential to pivot your build, and the only meaningful choice is made 5 minutes into the run, after which you're basically on rails. You still have to play well mechanically, but in order to achieve the roguelike power-scaling fantasy, almost all player agency is sacrificed.
For one last, double-barrelled example, I'm going to talk about two aspects of Pokemon. The first is during a story playthrough, where the Pokemon you catch earliest have the most time to level up, but this is mitigated by the Pokemon you encounter late-game being at a higher level so they can be slotted into your team without a need for catchup grinding (they miss out on EVs, but that's a level of optimisation that's completely unnecessary for a playthrough and in fact probably not even something most casual players know about). The Johto games (Gold, Silver, and Crystal, along with their remakes, Heartgold and Souldsilver) are infamous for having weirdly low-level Pokemon in their late game areas. Whatever your team is, it has to be pretty much decided by the midgame, as everything you encounter past that point will be too low-levelled (without some serious tedious grinding) to hold its own if added to your team. (To any Pokemon fans reading, this is not Johto slander - Johto is my favourite region). Whereas the previous examples above have pretty much every aspect of the game scale with the player, the Johto games inadvertently remove a key part of player decision-making by just forgetting to tune some scaling.
The other Pokemon example is the only other case (out of games I play) I can think of like Civ where a game can be decided before it's over and playing it out would just be a formality - competitive Pokemon. In a 4v4 match, it's entirely possible (and even somewhat common) to have a catastrophic early turn to the point where the game is just mathematically no longer winnable for you. For this reason, there's an option to just forfeit. The difference is that a match of VGC Pokemon lasts around 5 minutes, maybe 10 at most, whereas Civ takes several hours.
So what's the conclusion of this giant wall of text that really jnust amounted to me rambling about games because I think stuff like this is interesting? Essentially, snowballing is pretty inherent to games, but depending on the game and the genre, it manifests differently and can present different upsides and downsides. Civ (and other 4X games) is relatively unique in that options to scale enemies with the player are limited, but I don't think that makes snowballing a problem inherently. What I didn't mention about the Johto Pokemon games was that the enemy level curve is also a little wonky, and many of the late game bosses have sub-par teams. This is a FAR less common complaint than the wild Pokemon levels, however. Because most people aren't playing Pokemon for a gruelling test of strategy. They just want to make a team of creatures they like and have fun using them. Which is why the wild Pokemon levels come under such scrutiny, but Petrel's team of five level 30 Koffing doesn't. Snowballing will only ever be a problem when it impacts on the player having fun.
What that idea of player fun looks like varies, of course - in the roguelike examples, snowballing is a key part of the fun - but in civ I think the general experience most players are looking for is the opportunity to build an empire and manage it through various challenges over the course of history. The problem snowballing poses on that experience is that at a certain point, the challenges cease to exist. The lategame difficulty is an incredibly delicate balance. You don't want to apply debuffs and maluses to frontrunners and have players feel punished for doing well, as has been discussed, but I'd wager that scaling the AI on a curve so they become more threatening as the game goes on would also be unpopular. I suspect most civ players probably want to win the majority of their games. Civ games - unlike Pokemon VGC - are very long, and consistently falling at the final hurdle of an hours-long game as the AI boosts kick in would feel pretty frustrating for most people. Yes, there should be challenge, but the question should be how they overcome it and emerge victorious, not if they'll succeed.
I don't think snowballing is inherently a problem for civ. I just think that how it's thought about and addressed needs to fit the game and its genre. The problem snowballing creates is that endgames become meaningless and therefore boring. So how can that be dealt with to keep the experience players want intact? To me, having hit that critical mass to secure a win isn't inherently an issue, so long as I'm still engaged. The removal of micro helps a lot with this, as does my far greater enjoyment of combat. Even if my empire is strong enough by modern that I could beat the AI to any win I choose, weighing up alliances and ideologies, managing my troops to take the right settlements - that makes for engaging gameplay and I feel like I'm being presented a problem by the game and solving it. It's the polar opposite to the sit-and-wait tourism win in 6. So now they just need to make the other wins equally engaging for people who aren't lunatic warmongers!
(This got waaayyyyyy longer than intended and is especially overkill considering this thread isn't even about snowballing but in case it wasn't apparent the frustrated game designer in me likes talking about this stuff and hopefully someone finds it interesting lol)
