You'll constantly get into arguments about why we're not deploying the new upcoming thing that obviously fixes the issue.
At some point you'll need to switch over anyways, its better to get production battle experience by trying the new tech early and rolling it out where it makes sense.
It avoids internal factions and the "shadow infrastructure" that was mentioned.
Problem is, the old thing has known benefits and known flaws, while the new thing has known benefits but unknown flaws.
I think I'd claim the new thing has _claimed_ benefits. Very often those benefits turn out to be nonsense, driven by marketing and hype cycles.
By a Tesla, it has Auto Pilot!
By autopilot we just mean some slightly more advanced driver assist, it's not actually autopilot, and if you treat it as such you're probably going to die, at best.
The advice is that you take the new tech and do a boring thing with it first, like deploy or build something non-critical and gain the production experience that way.
The least interesting thing to do would still be to stabilise $old while carefully getting $new ready.