My recollection is of a deep irritation, like, "You people have me dorking around with all these hopeless special cases, when all along we could have been doing this? For fuck's sake, why?"
The action integral isn't solvable. You need to go to the special cases to get anything done. It is like saying energy is always constant. Sounds profound, but actually computationally useless on its own, since the invariant is broken frequently, until we go find something else to also call energy (gravitational potential, kinetic, electric potential, heat, mass! , etc)