It is for a Live Link between Blender and Daz Studio. Right now I use a local socket to transfer the data, but even with the localhost optimization, the data is still copied from one process to the other. So an already subdivided character model with mesh hair is huge and can still take a minute and a half per frame. With shared memory, there is no copying, and my tests (at least on Linux) show the transfer is virtually instantaneous (as it should be).
I am a professional developer as well, and so your point about the bleeding edge versus stability is well received; not knowing the dev’s position on that same point, I decided to ask and was prepared for something other than the answer I wanted I’m rather new to Python, but if it follows the strict Unix-y convention for version numbers (and I see so much of Unix in Python), a minor version upgrade should not break compatibility, making trivial whatever upgrades should be necessary. I can already hear you saying “That’s a really nice theory, Donald…” but sometimes when the planets align, it does work that way in practice, too. Thanks for your thoughtful points.