Why three mirrors, not five, not ten
A short note on the size of the mirror set.
The obvious question about the current Nexus mirror set is why it sits at three rather than five or ten. The answer is that three is the point where the marginal value of another mirror falls below the marginal cost.
With one mirror, any single routing failure is a total outage for the reader. With two, you have a fallback but no third option if the fallback also has a bad guard. With three, you cover most single-outage scenarios cleanly and almost all double-outage scenarios by chance. Adding a fourth mirror covers only rare triple-outage scenarios, and each additional mirror after that covers vanishingly rare compound scenarios.
The cost side is not zero. More mirrors mean more surface for phishing clones to blend into. More entries the reader has to eyeball against the captcha URL match. More work for editors verifying signed rotations. More cognitive load for the reader deciding which mirror to try first.
Three is the small number that trades off well. The operator has been at three for a while. Occasionally the set has drifted to four during a rotation transition (new address added before old one retired), which is a normal shape. It has not stayed above three for extended periods, which is a deliberate choice.