That performance regression only hit code that was using a very large number of paths with the same table of computed gotos at the end. That's likely to only be

That performance regression only hit code that was using a very large number of paths with the same table of computed gotos at the end. That's likely to only be relatively complex interpreters that were affected. So it's not a broad performance improvement. But it is nice to have an example of the compiler's new heuristic failing to prove evidence it needs to be tunable.