I work in storage, and this comes up a lot. I say "It won't reliably hit that target", and people think they know better. Then it works a good chunk of the tim

I work in storage, and this comes up a lot. I say "It won't reliably hit that target", and people think they know better. Then it works a good chunk of the time, and when it fails most of the time, nobody is paying attention. But one day, it doesn't just nip over the line, it hits the wall, and it fails hard in production. This isn't because I'm bad at making arguments, this is because every software developer has seen a few seconds where their storage device could hit X in a benchmark that isn't applicable, but it confirms their mental model of how this device works. It takes a long time to actually test these devices well enough to SHOW someone how this will fail, and they'll usually argue with you about how you're wrong for weeks. Better to just let them note your objections on the record, hit the wall with their face, and then point out your objections were well noted.