No, one should limit the level of abstraction to the fastest one that is reliable enough to get the job done. There's nothing intrinsic about high abstraction layers that say they must suck for debugging what you need debugged. It just depends on what you need to do with it. If it gets your job done faster and it's reliable enough for your job, then why not? That's why higher abstractions are created in the first place.

In short, you can't instantly rule out this compiler without knowing exactly what you'd want to do with it. And exactly how good it would be at that specific problem.