The function FUNCTION skip_update() and TRIGGER Dummy_Test_dupe really only prevent the UPDATE from happening, they can be omitted - I haven't yet tested whether it's more performant to have the TRIGGER or the UPDATE fire.

Interestingly, it made a small but consistent difference. I imported the same dataset (73613 INSERTs) three times into a previously populated database, and without the TRIGGER, i.e. with the UPDATE, it took an average 91.540s, and with the TRIGGER, i.e. preventing the UPDATE, it took 99.368s (each of these is only +/- about 2s), i.e. a difference of roughly 0.1ms per record. Not a big deal here, and not a rigorous test, but still interesting.

Update: Importing into a previously empty database revealed only a tiny difference: overall the import was roughly 10sec faster, again UPDATE beating TRIGGER with a difference of 0.17ms per record. Again, not really significant.