This reminds me of the physics "practical experiment" we had to do as part of the Cambridge physics tripos (at the Cavendish lab). It was also about measuring

This reminds me of the physics "practical experiment" we had to do as part of the Cambridge physics tripos (at the Cavendish lab). It was also about measuring gravity, using a pendulum though I forget the exact details, but we then iterated the experimental set up based on the initial results, in theory to get much higher final accuracy than our clock could give. I thought it was completely dishonest, because if your initial reading was off, then the final answer came out ludicrously wrong. The distribution of the outcome was no longer normally distributed around g, but now had a long tail of catastrophically wrong values. Presumably a good experimenter would simply have thrown away such an 'obviously wrong' value. I'm amused to think that a large number of those same physicists that happily sat through that experiment then went on to work as quants, where they essentially happily rebalanced probability distributions in exactly the same way, though with a lot more at stake this time around when they eventually got their catastrophically wrong value.