In NT, I will do it when I feel my partner does not have AKQ on that suit and I am holding Jx(x) (or even worst, in the case, you think there is a high chance of 3-3 break). The main theory on this is not to let the opponent gain anything from this suit given the high probability of betting a 3-3 break (in 5332/ 4333 distribution). Therefore, you may potentially gain 1 trick on another suit.
Another positive factor is that Suit is a major and is after a "weak" transfer and ends up getting an NT game. Therefore, you are betting the dummy is having A(Q)xxx, your partner has a K, and the declarer has (Q)Jx. But this is only worth it if you have another suit that needs to be established from your partner's lead.
If there is a trump suit, I will say I will only do it when I think there is a very high chance of ruff by my partner (e.g. I am holding xxxxxx on that suit, so the declarer should at least have a Q or J, hope that he/ she also has an x). The 3-3 break theory mentioned above for NT may sometimes work if you feel all players indeed have a balanced hand (from bidding etc.). However, my experience is it is usually better to find/ establish/ cash out your own suit first in this case.
In general, I think this works better in MP matches as +-1 trick is more influential. I personally seldom do fourth-best leads as I will always imagine (daydreaming) the best defense scenario (e.g. aim at cashing out my J in Jxxx).