Your experience is certainly encouraging, although I've yet to see a multi-point gain, myself, from a single encounter. I'm not sure I understand the distinction you are making between Locks & Traps and Pickpocketing, on the one hand, versus multiple Knock-Knock or Heal spells and Stealth training, on the other hand. Since the final patch, Pickpocketing, as I understand it, is in a class by itself. Similar to the way chest contents are set when a party first enters a particular map, so you cannot repeatedly re-open the same chest to try to generate that choice item (memories of the Gorrors' Chests), Pickpocketing results are also saved, so if you re-load to try again, you will still get exactly the same results. I don't know if that includes level increases. Someone will have to check that, sometime. (It should be pretty easy: just practice pickpocketing various NPCs until an encounter generates a skill increase. Then, reload that particular encounter and replay it to see if it leads to the same skill increase. One repetition might be luck, but if the skill increase occurs on every single re-load, it seems that this result is "set" like the other pickpocketing results.)
As for Locks & Traps, I recall others mentioning that they receive multiple increases by "examining" the same trap, or "picking" tumblers in the same lock, over and over. That, along with the Stealth, et al., increases, is something I've not been able to duplicate. What I'm wondering about is the mechanics of those - like you - who affirm it can be done. The one instance in which I HAVE seen multiple increases was during some early practice with the starting crabs on the monestary beach. But there, I got multiple increases because the combat would repeatedly stop when all the crabs were asleep and my party members were either defending or sleeping (the Bard) themselves, and then combat would immediately re-start as the crabs woke up. In between the stops and re-starts, my bard and ninja (the two characters training - Music and Stealth) would realize a one-point increase.
As for whether lots of practice (i.e., multiple attempts with low-level spells) versus difficult practice (i.e., fewer attempts with high-level spells) is more effective, I appreciate your opinion, but I've read a few contrary points of view. Obviously, I am coming at the question from a position of ignorance, so I'm not going to disparage the position of anyone with more experience. What I wonder, however, is what documentation or testing anyone has generated.
(I'd do it myself, really, but as Llevram rightly points out, I need to be playing the game . . . heh, heh, heh.)