Yes, I have brought a number of volume consumer products to market (Class II Metal Cased) as well as industrial street furniture (Junction controllers and associated items - awkward because despite being Class I in large metal enclosures, they also have large rubber sealed weatherproof access doors). In particular, the consumer stuff had to be tested in various operating modes where the case could be grounded (or not) by braided A/V cables, antenna lead etc.
Of course if you take something with a filtering scheme designed to work inside a metal enclosure and operate it outside that enclosure, it will likely fail on radiated. That's rather outside the discussion though, isn't it. The product or system that you take to the test house either includes a metal enclosure, or it doesn't (in which case it has no business being there).
It sounds as if you, or at least the test house, did investigate and (hopefully) understand why the led driver behaved differently with and without a mains earth. To me, it would have been a worst case scenario, a product that behaves differently in different installation situations (length and RF impedance of mains earth connection), something that requires an explanation in order to achieve consistent compliance in the field. Test house staff tend to have good intuition and knowledge from the range or products that they test.
The above sounds like an excellent example of where an unexpected external influence, in this case the variable effect of a mains earth wire of some RF impedance, is a clue to the underlying problem with the design of the product, not a cure - Something that must resolved to achieve consistent performance in the field.
As previously discussed, the mains earth connection is there for electrical safety, where required on Class I products. If it is influencing the EMC performance of a product then it points to an underlying problem that definitely needs to be investigated and resolved.
Edit: Sorry, that could have been written more concisely.