It seems like it'd be possible to automate gem testing for this scenario. Take the 1000 most popular gems, run the tests that come with the gem in 1.9.3. Run the tests again with the RC and see if any that succeed with the stable version fail with the RC.
Many Rubyists use Travis to do CI with their gems, and have ruby-head testing, often with allow-failures set. Here's one of mine: https://travis-ci.org/drapergem/draper