My sense is that the latest point release of the kernel is a better jumping-on point for development, rather than from -current, which is volatile & changing mu

My sense is that the latest point release of the kernel is a better jumping-on point for development, rather than from -current, which is volatile & changing multiple times daily. Perhaps when they start the branch, it's expected to be a small fix, & that it won't last very long. Making it work & making sure it won't break anything will be quick. So you'd try to keep your merge costs down by starting from the most recent version. It's optimizing based on faulty information. Building on a dev kernel might be really hard to troubleshoot, & merging is easier than that.