When that works, it works well. When that fails (e.g. people disagree as to how they performed relative to their peers), it fails badly. Performance reviews (if done well) lower the risk of system breakdown.
(Not to defend performance reviews explicitly--there may be a much better approach that accomplishes the same goal)
That's not necessarily true. When managers are doing a good job of regularly communicating with an employee and having critical conversations in regular 1:1s, nothing in a performance review should come as a surprise. If you've ever been surprised in a performance review, your manager failed in their responsibility--and you probably also failed yourself in that you didn't push for regular conversations.
Sure, those disagreements still happen. But, when things are transparent, there's very little risk of a "system breakdown" and a very high chance that a dissatisfied employee (who doesn't seem to be performing well in our scenario) will leave to make room for someone with more potential.
Edit: I just noticed the GP was jedberg; we both spent enough time at Netflix to see what each of us is describing work well in practice. :)
If scheduled infrequent conversations about that can adjust those expectations, wouldn't consistent frequent conversation adjust them sooner/easier?
In perfect world I'd have a mind-reading powered live dashboard I could check that shows a graph of anonymized averages of my perceived performance across broad categories (teams, peers vs subordinates vs leadership, etc.) and a single point for my direct report.
Getting feedback each year when there's no surprises feels like a formality and waste of time. In the rare case that there are any surprises it's disappointing and I get to make adjustments months late, having been left in the dark.
(Not to defend performance reviews explicitly--there may be a much better approach that accomplishes the same goal)