Some types of tests could be safely ran in additional, secondary CI workflows that could be executed, say, nightly.

One example of this test is the one I just created in #1280 that checks for a correct sampling distribution. In order for the test to be statistically meaningful, it has to use a lot of samples and it can be very slow, but we don't need it to run for all PRs, all the time.

Other examples of tests/scripts are big demos, stress tests, gas estimations, etc.