@tonton81 - There's no need to get upset. I see this mainly as a brainstorming session. We have this conversation in the lead-up to every new Teensy model.
Please let me explain how the Teensy panels are designed for manufacturing. Hopefully a picture will help.
View attachment 12379
Now the panels don't necessarily *have* to be made this way, but this shape is the result of years of gradually refining things with the contract manufacturers we use (PJRC doesn't have its own pick & place or reflow machines).
The left and right side of each Teensy are made by routed slots in the panel. The top and bottom are V-score in the panel, which a machine uses to cut the panel apart after it's fully built. V-score can't be used along the edge with the USB connector, because it overhangs the edge of the PCB.
The rigidity of the panel is a big issue for manufacturability. The pick & place machine moves very fast. If the panel flexes or vibrates, they have to run slower or risk parts shaking loose or off-center. Years ago this was an issue when we had longer slots. As you can see in the photo, we redesigned the panels to use short slots. A big unused strip in the center of the panel acts as a stiffening bar. When we added that, it dramatically reduced the vibration during assembly. It's one of many little refinements we've made over the years.
Another little feature you might notice in the picture are the 6 tooling holes in the edge of the panel. Currently we used only 2 of these. The contract manufacturer has a jig that was specially made for Teensy which pins that fit those 2 holes, and a spring loaded horizontal clamp on the other side. Without such a jig, they need to carefully align each panel to the solder paste screen. The jig lets them do this alignment once per run, and they can put extra time & effort into getting it really accurate if it's done only once. Then each panel gets placed quickly and easily with excellent repeatability thanks to the jig.
Just because we have a very well refined process doesn't necessarily mean we're forever locked into doing things just one way. But I think anyone can see how requiring one or both of the long sides to have routed slots rather than V-score is a huge change to the way we do the manufacturing!
A few years ago when Particle Photon (then "Spark") made their board with castellated edges, everyone was asking for this feature. But it requires using slots the define nearly all edges of the board, with "mouse bite" tabs. I saw those guys at a Hackaday event and asked if they'd had issues with the panel rigidity. They didn't answer with any specifics, but it was pretty clear that had been a big pain point. They wouldn't say what they'd done to deal with the issue. Maybe they've made some sort of carrier to secure it during pick & place? Maybe their panels are much smaller? Or maybe they've just tuned their placement process (likely running slowly) and just accept a higher failure rate or more rework?
On a personal level, I like to spend my time focusing on developing the software side and helping people get their projects working. Documentation is also a big goal, though you might not always know that from the sorry state of many of the web pages. Still, I *want* to work much more on that side.
Experience has shown that manufacturing problems absolutely kill any hope for software development & documentation, and even support suffers pretty badly. It's also incredibly stressful. Maybe this would be difference it PJRC was a huge company, but the reality is we're just 4 people and a couple contract manufacturers. Even with things as well refined as we have them now, problems happen all the time. It's just a fact of life in manufacturing anything. For example, just a few days ago we had a high failure rate in part of a batch. The bed-of-nails test just gives a pass vs fail. To actually diagnose *why* they failed requires quite a lot of my attention. Let me tell you, this has a way of suddenly becoming the very most urgent thing when we have customers waiting! In the case last week, it turned out a few of the PCB panels with marked defects (every panel gets electrical test at the PCB manufacturer) got built by mistake. Nobody noticed the remnants of the Sharpie X marks. I didn't even see them while handling the boards, until after I'd gotten info from the tester and was looking around the boards under the microscope! But ultra-urgent problems aren't always a quality issues. Pins starting to wear out and sticking on the bed-of-nails testers are another one that comes up occasionally. But just as often, something comes up that's entirely new and requires "drop everything" immediate attention to solve, if we're going to get everyone's orders shipped.
So while I'm not saying absolutely no to ideas like castellated edges, I do have a very strong desire to stay with this well tested panel design. Getting another jig made isn't too bad, under $1000 as I recall. But a panel using almost all routed slots with only small mouse bit tabs to hold the boards is a direction PJRC almost certain will not go.