Yesterday I said the test was the bad case, not the clean one. Today Priya ran the bad case, and the first half of it is in.
Start with the good half. The everyday congested home - 5 GHz, a video call running in the next room, the kind of traffic most people actually have - Alpha held a straight line. The worst round-trip in a thousand stayed under 80 ms, the single worst packet all run was 96 ms, and there were no dropouts and no holes in the data. Here's the honest catch, because we said we'd always state it: 62 ms is the safety cutoff we set ourselves, and a handful of those late round-trips ran past it. Every packet still came back and nothing went missing - so it's thin headroom, not slack. We're still working out exactly what that means for the cutoff, and that's part of Friday's write-up. For the common case, this is close to the answer we wanted, with the asterisk said out loud.
Now the half that earned the headline. Back on the 17th we caught a spike we couldn't explain and started calling it the ghost. Today Priya cornered it. On 2.4 GHz, with a microwave running, the link falls over on a schedule - a spike every two seconds, right on the magnetron's beat, and a real hole in the data each time. She instrumented it until there was no doubt. Her words: the microwave wasn't noise, it was the microwave.
I'm not going to print a hard number for the microwave case yet, because we've only got partial repeats and the decimal isn't settled - the full set runs tomorrow. But the shape is solid, and the finding stands: on 2.4 GHz, your coffee reheat owns the robot for as long as it runs.
That's the test doing its job. We have a named, repeatable enemy now instead of a ghost, and it's already half the argument for which band the real dock should live on. Friday I'll write up the whole thing, microwave tail and all, once tomorrow's run fills in the number.
- Callum