Trouble is, this ability is very hard to test for. A test is always an isolated, well defined problem, where the solution is written in a vacuum. But this is

Trouble is, this ability is very hard to test for. A test is always an isolated, well defined problem, where the solution is written in a vacuum. But this is completely unrepresentative of the actual kind of programming tasks at work (in my experience). Modifying the behaviour of existing code, that you didn't write, in a way that minimises technical debt, is a completely different challenge.