Alex Yermolinsky, in The Road to Chess Improvement, writes about the role of intuition in deciding when to sacrifice pieces --
“I asked Alex Shabalov what criteria, if any, he uses when he customarily sends his games into wild spins of tactical mêlée. He said that his main concern lies in a variety of ideas present in the positions he can reach in his calculations. If he feels it grow, then it's a good sign and he can go on; if it begins to diminish with every move, then the warning light comes on, hopefully before Shabba has gone too far with his sacrificial strategy.”
This reminds me of storytelling. By killing off an important character, a writer can sometimes open up an energetic field of alternative plot lines, transforming the relationships between all the surviving characters, permitting the writer and hopefully the reader a pleasurable rush of recalculation. An example that comes to mind is the moment in George R. R. Martin's A Game of Thrones when Ned Stark gets killed.
For me, later volumes in the series A Song of Ice and Fire seem to illustrate the converse principle -- that sacrificing too many valuable pieces/ characters too early in the game may be counter-productive. But since the final volume in Martin's series has yet to appear, I'll reserve judgment until after the endgame.
While we're on the subject of chess, I can't resist this quote from Vladimir Nabokov's Poems and Problems -- “Chess problems demand from the composer the same virtues that characterize all worthwhile art: originality, invention, conciseness, harmony, complexity, and splendid insincerity.”