Well, it would have cost them something and maybe meant a loss of focus on Ryzen, however against that if they had made a push with Bristol Ridge for DIY'ers the state and supply of AM4 motherboards might have been a lot better when Ryzen launched.
Both yes and no. Since AM4 has been launched since late september for OEMs, the platform was already in the wild for 5 months prior to the Ryzen launch. The kinks would have been worked out of the platform itself since, if not before, then. No OEM worth their warranty would have considered using a half-baked unstable platform.
The specific AGESA microcode for Ryzen is another matter, it would have been very much in development by september. Whereas Excavator microcode was pretty much a done deal once Carrizo launched in '15. I don't think it required that much effort to port it to the AM4 platform, since that capability was built-in from the start. In other words, its Ryzen that was not really "ready" by the time of its launch. The AM4 platform itself is/was fine.
BTW Ryzen itself is fine, even with a "beta" BIOS it is rock stable. It is only tweaking that is a problem currently. Particularly with regards for memory. If you run at stock, the platform is fine.
Another reason would be the simple lack of volume in the DIY space. The interest in APUs is slim at best, and AMD already had Steamroller APUs on FM2+ for that market. There really isn't much difference between Steamroller and Excavator. Both have their strong points and weaknesses, while AMD would love to sell every one of Carrizo/Bristol Ridge dies to OEMs.