This is incorrect. Einsteins theories were based on theoretical considerations and probably not on direct observations at all -- although there is controversy about whether he knew about the Michelson-Morley null result -- there is very little doubt that the Special Theory of Relativity was based almost ENTIRELY on a consideration of the theoretical inconsistencies between Newtonian Physics and Maxwell's Equations ALONE.
The discrepancies were known for a few decades, certainly not for "millennia." But in any case, they had very little to do with experiment.
The constancy of the speed of light in vacuum is Einsteins second theoretical postulate of The Special Theory. Measurements of the speed of light in vacuum had not been going on "for millennia." And experimental physicists did not generally accept Einstein's explanation that c was a constant in all reference frames for quite some time.
The General Theory of Relativity also had nearly nothing to do with experimental observations. It explained the precession of the perehelion of Mercury nicely, but that was a very minor detail that didn't concern physicists terribly much. Einstein worked on the General Theory of Relativity as a necessary -- and purely theoretical -- generalization of the Special Theory of Relativity, which, as we have already seen, owed far more to theoretical consistency than to observation. The General Theory was not motivated by, and had virtually nothing to do with experiments, at all.
Neither did his explanation of the photoelectric effect, which was much more influenced by purely theoretical concerns pointed in his direction by Max Planck's resolution of the Ultraviolet Catastrophe through the quantization of black-body radiation. Planck had already laid the groundwork, and Einstein proposed quantization to make Plank's resolution and his own explanation of the photoelectric effect conceptually simple rather than ad hoc. Planck, in turn, was motivated to remove operational infinities from the mathematics of the theory of black-body radiation, which was already known. Again, a search for theoretical self-consistency; since the Stefan-Boltzmann law already gave the correct experimental temperature dependency of black bodies.
Those theories were stepping stones.
They were, as are all theories, but they were theoretical stepping stones, and not [for people who actually know the history of physics] experimental ones.
It is no different than when the theorists came up with the laws of thermodynamics.
This statement is false BOTH conceptually and historically. It could not possibly be MORE WRONG.
The Laws of Thermodynamics [except for the 0th law] are entirely observational, and have nothing to do with any theory. As a matter of fact, the classic post-graduate text on Thermodynamics [referenced at the end of my Home Page] says very clearly in the introduction: "If the atomic theory, the kinetic theory of gases, classical physics, statistical mechanics or quantum mechanics, or any other underlying theory of physics were invalidated tomorrow, it would have no effect whatsoever on the laws of Thermodynamics."
The laws of Thermodynamics are triumphs of experiment and observation, and they have no inherent theoretical basis whatsoever. ALL of them were originally stated in terms of observations made with heat engines.
Actually, and ironically both physics and thermodynamics share a very fundamental rule — matter and energy are not lost or disappear. All else is secondary.
Actually, and without the slightest hint of irony, this last statement has some howlingly ridiculous errors. Thermodynamics is a branch of physics. It's not separate from it. In terms of macroscopic phenomena, it must necessarily share that with which it is a smaller part. The claim that physics has a very fundamental rule that matter and energy are not lost or disappear is, on the quantum level, untrue. In quantum field theory virtual particles come into and go out of the vacuum all the time. What holds up is that if time translation invariance exists, an experiment will not measure the violation. But this is not the same thing.
In the the real world Experimentalists measure the expectation values of quantum observables. Those expectation values must imply the conservation of energy, because if they did not, we would also see observable violations of the law of energy conservation in thermodynamics... and we don't.
And there is nothing "ironic" about the requirement that a self-consistent theory must contain implications which are testable across domains.
As to the statement that "All else is secondary." Actually, this isn't the way physicists currently think about the law of conservation of energy. The 1st Law is actually of secondary importance to a specific symmetry, which we believe is exactly conserved. That symmetry is time-translation invariance. But even there, I doubt too many physicists would claim that to be the most important theoretical principle in physics, and NO physicists would agree that "everything else is secondary."
And absolutely yes, each theory was a stepping stone all the way back to the Greeks and the first definition of an atom, or Newton, or Faraday, Maxwell, Plank, Einstein, etc. etc.
As for thermodynamics and Physics and atom smashing and conservation of energy and matter:
The process is thus:
Fire a high speed particle, a proton?, with a relativistic mass and velocity at an atom and compute the starting energy. After impact, count up the resulting particles, their speeds, charges, etc. and compute energy and add them together.
THE ENERGY BEFORE BETTER MATCH THE ENERGY/MATTER AFTER. Sounds familiar. Matter is not destroyed?
This is not unlike when physicists of Einsteins time were trying to explain missing mass during a fission experiment. Einstein proposed that the matter was transformed into energy. It was destroyed but transformed.