Here's how Lewis and Short put it:
dē-curro, cŭcurri or curri (cf.: decucurrit, Caes. B. G. 2, 21; Tac. A. 2, 7; Suet. Ner. 11: decucurrerunt, Caes. B. G. 2, 19, 7; Petr. 64, 3: decucurrerat, Liv. 1, 12: decucurrisse, id. 25, 17; also, decurrerunt, id. 26, 51; 38, 8: decurrēre, Verg. A. 4, 153; 11, 189: decurrisset, Liv. 33, 26), cursum, 3,
Both forms, in other words, exist, and the reduplicated form seems to be classical. In Vergil, the unreduplicated form might be avoided metri causa, since decucurr- forms a cretic that doesn't fit his dactyls. Livius uses both forms, perhaps just because ... they both existed?
EDIT: Grammar is a descriptive science: if something exists and makes sense, we have to note it and accept it. It's sad but true, I think we would all prefer to have less forms to think about when learning a language.
EDIT EDIT: I think reduplication in general had a hard time in Latin: consider tuli, the perfect of fero, which originally was reduplicated ( = tetuli, you can find vestiges of it in the two t's of rettulit from refero), or consider spondeo, which sometimes loses its reduplication. There are probably other and better examples.