The board inside the K2 went through nine revisions. Six of them were wrong, and all nine are published.
Why publish failures
Because the interesting information is in them. Revision four had a ground plane split under the crystal that produced an intermittent enumeration fault on about one board in forty — the kind of thing that ships and then generates a year of returns nobody can reproduce.
What shipped
Revision nine: single-sided assembly, one crystal, socketed controller, and a USB-C daughterboard on a ribbon so the port can be replaced without touching anything else.
