"I don't feel ready to commit to having the chips pre-programmed" Why not? This is no effort at all, you just contact a company dealing with such parts, give the binary to them (NDA if necessary) and buy the pre-programmed MCUs from them instead of the silicon vendor. There's usually no big MOQ.

Also, justifying development costs of pogo pins, bootloaders etc means that you must have large volumes. Engineers cost money. How many 2x5 1.27mm SWD connectors can you buy for the engineer salary required to develop this? Almost certainly far more than a product life time of such connectors.

@Lundin , there's a lot of truth to that. At this scale, a streamlined programmer is a "nice-to-have". I'm trying to think ahead, get the technique down, etc. Having the ability to flash it in-house (even if pre-programmed) is valuable no matter what. We do lots of product variations and connecting to each device by serial-over-USB is a pain. The bootloader was a must-have for field upgrade, and to be honest was kindof trivial to develop.

The way I usually do this is to buy pre-programmed MCUs for high volume products ~1000pcs/year, keep the SWD connector footprint on the board but don't mount it. For products of lower volumes or products where you expect a lot of software releases, keep the connector and use an in-circuit programmer. As for bootloaders, they are indeed much easier to make nowadays since most chips support making them in some way. But they are for specialized use-cases only.