Follows on from #464 and #465.

We've added some specific tests for the Settings object's equality behaviour, and have reached some minimum code coverage, but this is really something that hypothesis should be able to help us with.

In particular, it should be reasonably possible to construct a Hypothesis strategy that will produce Settings objects that we can then evaluate. This is not an easy thing to do as it requires some familiarity with Hypothesis to pull off, but is probably a worthwhile challenge for anyone who wants to get their hands dirty with Hypothesis.