With factories, data can be defined on attributes in a static way, or the fuzzy functions can be utilised.

It has been asserted by some, that the fuzzy functions should be generally avoided, and static values should be used to avoid potential flakiness - if the values differ from test run to test run, there is the potential that one (potentially rare) test run will generate an invalid value, but it may be hard to reproduce and diagnose

We should document approaches for this - e.g.: when (if ever) to use the random generation utils