Who Is Trapping Whom?

MY round-six game today from the Highlands Open in Trebic - part of the Czech Tour - featured a move by White in the Hyper-Accelerated Dragon that at first sight looks like a blunder … but it has been played by strong players, including Lothar Vogt (2515).
Jindrich Habada (1761) - Spanton (1923)
1.e4 c5 2.Nf3 g6 3.d4 cxd4 4.Qxd4 Nf6 5.e5 Nc6 6.Qh4?!
This move probably deserves a straight question-mark, but there are 74 examples of it being played in ChessBase's 2018 Mega database, and White scores a not-so-dreadful 44 percent.

Black to play and win material

6...Nxe5
Felipe de Cresce El Debs (2531) preferred 6...Nd5?!
7.Nxe5 Qa5+ 8.Nc3 Qxe5+ 9.Be2

How would you assess White's compensation?

Black has won a pawn but White has a slight lead in development and the prospect of good play for his rooks on the central files.
Perhaps it is not so surprising that the analysis engine Komodo9 reckons Black's advantage is not quite worth a third of a pawn. Stockfish9, on the other hand, has Black almost two-thirds of a pawn ahead.
9...Bg7 10.Bf4
Vogt (2515) - Markus Schaefer (2270), Bad Woerishofen 1992, saw 10.Bh6 0-0 11.Bxg7 Kxg7 12.0-0-0, with chances for both sides, but it is hard to believe Black is not better (½-½, 53 moves).
Also sensible looking is 10.0-0 0-0 11.Bg5 d6 12.Rfe1, and White went on to win in John Dodgson (2135) - Spanton (2010), Paignton 2008.
10...Qa5 11.0-0 0-0 12.Bh6
Viktor Ianov (2360) - Alexander Vaulin (2510), Policka 1995, saw 12.Bf3 d6 13.Rfe1 Be6, when Black is clearly better, but the game was drawn in 47 moves.
12...d6 13.Rae1 Be6 14.Bd3?! Qh5 (0-1, 31 moves)
So, to get back to my original question, who is getting trapped? I guess the answer has to be White, but there is no doubt White gets some compensation, and practical chances, even if most Whites, I suspect, play 6.Qh4?! without realising it loses a pawn.
In conclusion, then, annotating White's sixth move as "dubious" rather than "bad" may well be right after all.