I'm currently reading (and enjoying) Van Perlo's Endgame Tactics. Recently I reached position 341
The game continuation was 1... e3 2. Rd6+ Kc3 3. Re6 Kd4 4. Rd6+ Ke5 5. Rd8 e2 6. Re8+ Kf4 7. Rf8+ Ke3 8. Re8+ Kf2 9. a7 Rh1+ 10. Kg5 Ra1 11. Rf8+ Kg3 12. Re8 Ra5+ 13. Kg6 Kf3 14. g5, at which point van Perlo says "Checks no longer help as after 14. Rf8+ Ke4 15. Re8+ the rook can be interposed".
This is, of course, a disastrous suggestion: after 15. Re8+ we reach this position:
when there can hardly be enough question marks for 15. ... Re5, allowing 16. a8=Q+. (Black does in fact win after 14. Rf8+ by 14. ... Kxg4, so the game at least is not spoiled.)
I wonder how this error slipped through? I'm reading the book's second edition, so it has presumably survived unnoticed (by the publishers at least) for quite some time. The preface talks at length about how the positions have been computer checked; but apparently not completely thoroughly. My guess would be that the fatal mistake was leaving "the rook can be interposed" in words, rather than entering the move 15. ... Re5. The computer isn't ready to analyze chess in a natural language just yet.
I'm not sure what more to say about this, except that I found it rather an unnerving experience to see such an error in black and white, right there in front of me. A blunder like this shakes one's confidence in a book more than a little.