Miscelaneous travis test failiures

This issue is a bucket for test failures which have occurred on the travis tests, probably related to the fact that the tests are run in parallel.

I'm not unduly worried by these failures - it is something I'd like to fix, but not something which needs to hold up a release.

The problems will inevitably prove hard to reproduce and will be dependent on test execution order (which changes with hash ordering and performance of particular nodes).