What has an online input tester told you that turned out to be wrong?

I'm launching Saturday, and this product exists because of wrong numbers. Most browser-based mouse testers count one movement per displayed frame, which means they're measuring your monitor (a 1000 Hz mouse reads as 60 Hz on a 60 Hz display), printed with total confidence.

When I started measuring properly, I ended up publishing error bands instead: every number on InputProbe comes with the accuracy band it was measured to, and when a capture can't support a number, the tool refuses to print one and says why.

Before launch I'm collecting failure stories. Have you ever used an online DPI checker, polling rate tester, double-click tester, or keyboard tester and gotten a result that didn't survive a second look (a DPI that didn't match reality, a polling rate that changed between runs, a "faulty switch" verdict on a healthy mouse)? Which tool, and what did it tell you?

I'll take the failure modes people report here, reproduce them, and write up what actually causes each one.