The industry this was in was moderately high pressure, but by no means a grind like you might see in games, etc. We painstakingly avoided bs deadlines, for example.

To a first order approximation, five years by 200 engineers is a thousand engineer years. So naively, that'd be two PIPs per thousand engineers per year. Of course, you can argue that this isn't true since the performance of an individual from one year to the next is correlated. But this was a very rapidly changing organisation where people move between departments regularly and there's a lot of growth. The latter means that you get a lot of new people that might turn out to be a poor fit. So I actually think the original approximation of quite conservative: 2 out of 1000 engineers being disagreeable about their performance assessment sounds like a total luxury for managers!

But you're right to point out that that's only the bad cases. Our company had a(n HR imposed) rule that anyone receiving sub par ratings (needs improvement) for two consecutive quarters had to go on a PIP.

In a growing organisation, complexity keeps increasing: more coordination overhead. More things all going on at the same time, more scaling challenges, etc. So naturally, performance that was just about alright last year just about DOESN'T cut it this year. This means that in my experience, it would be a sign of rating inflation or managers that avoid difficult conversations if the fraction of "needs improvement" in any given quarter was less than ~5 percent. That's a number I just pulled out of thin air, but in the right ballpark in my experience. And there you go: you get several times as many PIPs as the "bad" ones discussed above.