I think there's an aspect here that is overlooked. The tests are designed to be carried out in order to safely find faults. They are really designed for new build, ie everything starts dead. The tests don't work so well for fault finding and periodic testing, which is why 2395 is one of the highest qualifications. I am on record elsewhere on this forum for stating if you are Part P only you are not qualified for periodics or for offering a fault finding service.
in my view, the OP has made an early assumption which was wrong, and nearly killed himself. Also his powering down of the CU was at fault.
assumption: rcd faulty and was not. This was replaced and not tested. A functional test of the new one to prove the assumption would have shown the assumption wrong and further investigation was required. However he assumed fault fixed and ploughed into a full test of a system which wasn't dead. Golden rule: change one thing at time to prove cause and effect.
powering down CU. OP States no data. It would be better to identify what was powered down by each MCB as it was switched off. Ie identify the circuit. The false feed would be identified as Unknown which should equal warning. With an apprentice to run round the cafe and see what's happening this form of investigation is easy.
golden rule. Take your time.
If you think this is easy to say sitting on the computer, I had exactly this last week. Unlabelled CU with additional circuits added in. I spent 2 hours finding out what was on each MCB and where other supplies in the installation interacted before even thinking if putting my fingers on the copper.