I had unit testing jobs running ALL tests in source (full sandbox) and target (production) org - resulting in 80% and 79% test coverage. However, when I started to deploy my changes - the validation failed because test coverage was only 74%. Finally, after a few days, I figured it out.

I was just simply clicking "Validate" (i.e., did not choose specific tests). In the past - that ran "all tests". But something seems to have changed. When I picked "choose tests", about half the tests in my org were "automatically selected". I manually selected all other tests - effectively forcing all tests to run. This time the validation passed!

It would be great if there was some output from Gearset during the validation in terms of what tests were being run (partial vs all)