Your example is pretty illuminating. AAA games are one-and-done: you ship, and then the codebase is basically frozen, and you make money off expansions. The other area where team sizes top out at 50-100 people is desktop software, where (barring future versions) it's also one-and-done. You sell a software package, money comes in, that's it.

The way the calculus works for Internet-based SaaS is: say you have revenues of $4B (like Shopify), and an engineer costs $400K. If the engineer works on a project that improves sales by 0.01%, it's profitable. If a team of 10 engineers works on a project that improves sales by 0.1%, or 100 engineers improving sales by 1%, it's profitable. How many 0.1% opportunities do you see in the product? How many 1% opportunities? What if you add risk? An engineer might fail to see any effect on 4 of their projects, but if they improve sales by 0.1% on the 5th, they've still made the company twice as much as they cost.

It's hill-climbing. There are usually lots of 0.1% opportunities available with a big worldwide SaaS. Things like localization and entering new markets will generate way more than that; even things like moving around CTAs and changing colors will usually generate more in profit than the engineer costs.

Also add to that diminishing marginal productivity of each engineer. In a team of 5 moving a button around takes an afternoon. In a team of 1000 it often takes a couple quarters, because you need to get sign-offs from all the teams that the code touches, run it through QA, and gather metrics to show you didn't regress anything. The engineer is still profitable, because the effect of their change multiplied across millions of users is way more than their salary costs. But they're not as profitable, because they work so slowly.