Where do I start?
Here's a relatively trivial exercise for a developer: Write a Cucumber test that implement the journey through CEST, then configure it with a number of test cases that have the answers that would have been provided for every case that has ever been through court. Cucumber is used extensively on government projects, I've been there and done that in at least two departments in the last three years. If you have already set up a Cucumber test framework, implemented some Cucumber tests (on pretty much any level) and have automated the running of those tests in CI/CD it is at most a week or two's work. Once. To have a set of test cases that can be run at will to demonstrate compliance with requirements.
To claim you've kept no evidence of testing is not credible. If a developer said there was no way he could prove his software met requirements to my face I'd laugh long and hard and then tell him to go back and do his job properly. Evidence of automated testing (at every level from unit through to smoke) from CI/CD is how modern projects are delivered. Including everything HMRC has done for at least the last three years.
For HMRC to say this, well... they're actually talking to the people who do this day in day out for a living. Credibility drops through the floor at this point (assuming it was above the floor to start with, hardly a given here). How exactly did they satisfy themselves that their subcontractor delivered a system that worked as they wanted it to? And still does? Total crap.
HMRC don't want evidence of testing because it would stop them repeating the complete rubbish that it provides "accurate results". After all, when it ignores mutuality of obligation how can it be accurate? And if they are forced to admit that the tool is flawed, and maybe even have to fix it, they've just had their enforcement baseball bat taken out of their hands.