Salt and Silicon
The tide came in at four and took the prototype with it. Not all of it — the enclosure was rated for splashes, not immersion, and the salt water found the one seam we hadn't potted. By the time I got down to the tide line the board was still blinking its status LED under an inch of foam, gamely trying to hold a Bluetooth connection to a phone that was, by then, thirty metres up the beach in my jacket pocket.
What survived
The battery survived. The LED survived. The little accelerometer that was supposed to log wave impacts survived, and logged one enormous, sustained impact that looked, on the graph afterward, exactly like what it was: a device tumbling down a beach in white water.
What didn't survive was the antenna trace, which had never liked being this close to a ground plane in the first place, and which chose this moment to finally let go of a solder joint that had been marginal since the first prototype run. In hindsight it had been marginal since before that — visible in retrospect in a hairline crack under the microscope, invisible at the bench, invisible right up until thirty seconds of salt water found it.
The rebuild
The second enclosure gets a full pour, not a splash rating. It also gets a tether, because the smartest waterproofing in the world doesn't help if the thing being waterproofed is now somewhere off Portugal. Silicon is patient. Salt water is more patient. The trick, it turns out, is mostly just not needing to find that out the hard way twice.
If you build anything that goes near the sea, assume the sea has read the datasheet more carefully than you have, and budget the tether before you budget the sensor.