Tests checking the metric are unhelpful

There are quite a lot of tests that check that metric(kernel) returns a particular metric. It's not clear to me how much value these tests have.

It probably makes sense to ensure that SimpleKernels do implement metric, and return a valid type of metric, but I don't think that we should be testing that they return a specific metric, as this feels like writing a test is basically the same code as the code it's testing, which I believe generally gives a false sense of security.

I'm pro- testing that kernels yield the correct numerical values for certain specific inputs / parameters, or that they are the same as other kernels for particular settings (e.g. gamma-exponential and EQ / exponential) as with this this type of test you're typically not just writing out your source code in your tests.

What are people's thoughts?

Search the test suite for checks that metric(kernel) returns a particular metric, then review the SimpleKernel and metric discussion in this issue. The work is done when the project has a decided, consistent testing approach that preserves numerical and equivalence checks while resolving the disputed specific-metric tests.