This sounds like a good idea for execution of small ideas but most innovative things are actually very fragile in the early stage. Its almost the opposite of failing fast for those type of projects that truly makes them successful, for example the ASML EUVL would have failed instantly given how "impractical" it was according to every major expert at the time. I think it is more useful to think of a balance between exploration and this type of fail-fast aproach.
