Hume’s real argument isn’t that miracles are impossible but that testimony to a miracle is never sufficient, since a mistaken or lying witness is always more probable than a suspended law of nature. Redefining “violation” doesn’t touch that.

Correct, and this is the half of Hume that the definitional reply does not reach. The standard response, from John Earman, Richard Swinburne, and others, is that Hume’s probabilistic argument, taken strictly, would rule out believing in any extraordinary event on testimony, including well-attested but rare natural events; and that the probability of a miracle on testimony depends on the prior probability that God exists and would act, which Hume simply set at zero. Earman’s Hume’s Abject Failure is the technical treatment. The evidential argument is the one to answer, and it is answered by the wider case for God, not by a definition.