In our tests that ensure we are able to transform Nighthawk's native output format to Fortio's,
we use rather arbitrary and large sample data, see input and the corresponding
output.

In #534 we are more or less forced to propagate this into yet a new gold file, which by implication, also needs to be large and arbitrary.

Long gold files make it hard for the reader to figure out which portions actually are relevant to tests and which are just noise. It would be good to could cut these down to the minimum required to test this the expected behaviours?

Context: #534 (comment)