In thinking of late about science, and how it isn't a prediction machine correlating expermintal measurements, I keep thinking of the notion of paradigms. It isn't perfectly clear how this is different on the one hand from really basic notions typical of the scientific enterprise, both narrowly and broadly construed. For example, the notions of causality; truth is correspondence to reality; nature is lawful (or if you insist on speaking unnaturally, nature displays some regularities to which we do not find exceptions and which heuristically used can make repeated valid predictions to the outcome of experiments); nature is consistent or coherent or intelligible (yes, we all know about quantum woo, but the paradox is that despite a theory that is set in a Hilbert space nonetheless we get results always compatible with good old Einsteinian spacetime.) Although not even these notions are held to be above revision, in practice they have been incorrigible because they are more inductions than deductions from theories that have had to be revised. If these notions are part of the scientific paradigm, which seems a valid construal of the notion of paradigm, then the scientific paradigm has never successfully been replaced. Merely (a giant word!) modified. In this sense, the notion of a succession of paradigms succeeding each other willy nilly is utterly refuted.
On the other hand, if paradigms refer to specific ideas, things like atoms; the distinction between energy and momentum; the field concept; the wave concept; the ether; phlogiston; the Ptolemaic model; spontaneous generation; Newtonian absolute time, then one thing that pops out from even this cursory list is that such paradigms are often not wholly overthrown. Consider atomic and wave and field conceptions of phenomena, all still live and kicking in one context or another. The coexistence of such broad differences in how scientists view things seems incompatible with the notion of incompatible paradigms that can only replace each other. But these notions are so general it is hard to see that you could just demote them to simple concepts.
Another thing that emerges is that the dubious element in these constructs, such as the centrality of the Earth or the doubtful coherence of a notion of absolute time apart from change, were usually challenged by alternatives. Consider Aristarchus' model or the objections to Newton's absolute time (and his theory of light, as well.) A one paradigm may be as extinct as the dinosaurs, replaced by a ratty little one that had, after the eon scurrying in the undergrowth, suddenly grows, multiplies and branches out. But it didn't come from nowhere.
And it has often been observed that the empirical observations of these theories, and the predictions they led to, though not the verbal forms and pictorial analogies, were conserved. The "paradigms" if that is what you wish to call such related sets of concepts that articulate fundamental theories (scientific explanations) of the world are connected to each other. They are not arbitrary. Of the few that are wholly obliterated by induction, a bare few have seemed to reemerge, notably spontaneous generation, and in a few advanced speculative models, the ether and absolute time. Perhaps there was an inspirational aspect (but for humans, inspiration can come from religion, philosophy, play, dreams, whatever serves to propel the speculative moment of science---but it isn't science if you don't try to find out if it's true.) Closer investigation reveals that chemical biogenesis is nothing like spontaneous generation. Etc.
Lastly, if a change in paradigm can be usefully labeled a scientific revolution, then the notion of counterrevolution really should be allowed. In biology, resistance to evolutionary science has plainly been fostered by a religious paradigm trying to turn back the clock. I think it is pretty obvious how reactionary weltanschaungen contribute to economic or psychology or history. Thus it appears that the paradigm of science as a random sequence of paradigms unconnected to each other is wrong. And that appearances of retrogression can be reasonably attributed to rather literal reaction.
In short, the dwellers in paradigms may be happy, but they are not useful.