Nothing for Critique to Stand On
Squash a repo's history down to one commit labeled "initial import" and you haven't just tidied the log. You've erased every place a reviewer could point to and say: right here, this is where it should have gone a different way.
That's the part worth sitting with. Incremental critique isn't "this is bad." It's "at this step, option B was sitting right there next to option A, and B would have gotten you further." It needs a sequence of actual decisions, each one made with alternatives on the table, to attach itself to. Collapse the sequence and the critique has no surface left to grip. You can still say the result is bad. You can't say where it went wrong, because "where" no longer exists.
This is why "arrived whole" is such a useful story to tell about a thing you made, whether or not anyone planned it that way. Not "we iterated through six approaches and this one won," but "this sprang out complete, armed, ready." The second version isn't a compressed retelling of the first. It's a different claim, one that happens to be immune to a whole category of objection. You can evaluate the finished thing on its own terms, sure, but you can't ask whether an earlier fork should have been taken instead, because the story insists there was no fork. There was one unbroken act, no tributaries, nothing upstream to second-guess.
The convenience doesn't require a conspiracy. Nobody needs to sit down and think "how do I armor this against process critique." The sprung-from-nowhere framing is just the more flattering story, and flattering stories get told by default. It reads as confidence, as inevitability, as a thing so right it didn't need to grope around first. The defensive side effect comes along for free. That's exactly what makes it durable: the people reaching for it usually aren't lying, they've just picked the version of events that happens to leave no joints for anyone to pry at.
Outcome critique still survives this. You can always say the current thing underperforms, costs too much, breaks under load, treats people badly. What dies is the more useful kind, the kind that produces a lesson for next time: at the point where you chose speed over durability, that's where it should have gone the other way, and here's how to recognize that fork when it comes around again. You can't extract that lesson from a story with no forks in it. The whole-arrival version isn't just unflattering-fact-free, it's lesson-free, which is a worse cost than it looks like from the outside, because it's the cost that compounds. The next iteration can't learn from a decision that officially never happened.
Worth noticing this works at any scale, not just the grand ones. The squashed commit is the smallest version of it: nothing dramatic, nobody deceiving anybody, just a convenient flattening that happens to also mean nobody can ever ask whether schema v1 should have looked different before it became load-bearing everywhere. The sequence existed. It's just not on the record anymore, which, for the purposes of critique, comes to the same thing as never having happened at all.
A thing that can't be shown a fork in its own path can't be told it took the wrong one.