We test IP67 in-house, every batch, by submerging samples in a tank for the full 30 minutes at one metre. Most batches pass without drama. In July 2024 one didn't, and it cost us a week.
The batch was a run of 300 rechargeable handheld work lights. We pulled the usual samples, dunked them, and two out of eight had water inside the battery compartment afterward. That is an automatic hold on the whole batch for us. You cannot ship one light that says IP67 if you've seen water get in, because the buyer will find the one bad unit, not the seven good ones.
When we opened them up the gasket looked fine, seated properly, no obvious gap. Took us a day and a half to find it. The problem was the gasket itself, not the design. Our usual supplier had been out of stock and the purchasing guy had taken a substitute from another vendor for that one run. The substitute rubber was very slightly harder, so it didn't compress enough when the housing screwed down. Half a millimetre, maybe less.
So we re-worked the lot. Every unit got the housing opened, the wrong gasket pulled, and the correct softer gasket fitted, then a re-test on a bigger sample, 30 out of 300 instead of the usual 8. All passed the second time.
The lesson wasn't really about the gasket. It was that we let a substitute material into a sealed product without re-testing it as if it were a new design. Now any change to a sealing part, even just a different batch from the same supplier, triggers a fresh dunk test before it goes to the line. Henry wrote that into the QC sheet himself.
It's not glamarous work, sitting there with a tank of water and a timer. But the whole value of an IP rating is that it's true. The day it stops being true on our boxes is the day a mining crew stops buying from us, and rightly so.
