This I always forget. High costs and the thread of litigation are the core thing killing GA, or at least keeping it from growing. But to that point, rather than a clean sheet design, why not adapt an automotive engine? I am sure it would require some non-insignificant updating, etc. but you would have years, even decades of engine reliability data so it is not really "clean sheet" design. Oh well. just spit balling here. Pardon the thread drift
Using automotive engines has its own issues as well. Austro did a better job, but the original Thielert engine had a lot of technical issues. In both cases it weighed more than the equivalent Lycoming, although without a doubt had a lot of benefits.
From a certification perspective, there are issues as well. The way auto makers build engines (and more importantly implement changes) doesn't sit well with the FAA, so you can't just pull engines off a production line. The way that Thielert got around this was by buying one day's worth of engines from Mercedes, and they certified around that one day, for which no changes were made. Problem was that when they ran out of engines, they ran out of engines. So there were longevity support issues built in from the factory. That would be avoided to some degree if you used something like a Ford 302 as your basis (which is what Ben Haas did) with aftermarket aluminum blocks/heads/etc. Then you could probably manage to minimize or eliminate the number of changes, and basically just take a proven automotive design to work with. But if you want to use a modern automotive engine that's going to get updated and changed, it's harder. This assumes you can even find an automotive OEM who will work with you.
Support wise, there are other challenges. A&Ps know how to work on Lycomings and Continentals, but that's about it. Airplanes are very mobile and tend to break down in inconvenient locations. So now you break down somewhere with your Uber Mach 5 engine, and the mechanic pops the cowls only to see a mess of hoses and wires and starts scratching his head. Oh, and he doesn't have the special whiz-bang-only-available-from-uber-code-reader (that'll be $10,000 for one of those), you get the idea.
Previous attempts have largely had issues because they either weren't developed well (this is typically the case in the experimentals that have them) or never matured. In some cases a bit of both (I think Thielert was a bit of both).
That's not to say I don't think the idea is great on the surface. I do, and 10 years ago I said I thought that would be the way of the future. I was right to some degree. Thielert was already around then and Austro came after them. What's left of Thielert is now owned by Continental. But in aviation "future" requires a lot of patience to get to. If I built an experimental, I would build it with an automotive derivative engine (well, plural, because I'd build a twin
). That said, you could also make some relatively minor changes to current aircraft engines and get some very significant improvements.