In the event of a Mode IB abort during an Apollo launch (an abort initiated between 3 and 30.5 km altitude), a pair of canards would be deployed from the tip of the spacecraft's LES in order to force the command module to weathervane blunt-end forward, because the CM's parachutes (stored at the apex of said module) were only designed to deploy downwind (with the CM's apex pointing backwards, and its blunt end facing forward).1
However, on 19 May 1965, during the A-003 abort test, the Little Joe II booster propelling the spacecraft suffered an almost immediate loss of control,2 and started to inexorably roll faster and faster until the launch vehicle disintegrated. This automatically triggered an abort, and the canards deployed, but, as the capsule was still rolling rapidly (at a rate of approximately 260 degrees per second, according to Wikipedia), they failed to flip the CM tail-first. As a result, the parachutes deployed into the wind... and worked perfectly, despite the nose-first attitude and extremely high roll rate.
Why weren't the LES canards deleted (which would have made the LES, and, thus, the vehicle as a whole, somewhat lighter), despite their having been proved unnecessary (as the parachutes turned out to be perfectly cool with deploying with the CM nose-first)?
1: For a pad abort, or a Mode IA abort (between launch and 3 km altitude), the canards would still deploy, but the job of flipping the CM ass-first would be performed mainly by the LES's pitch motor; for a Mode IC abort (between 30.5 km altitude and LES jettison at ~90 km altitude), the canards wouldn't have enough control authority to flip the CM (because thin air), and the CM's RCS thrusters would turn the trick (and the CM) instead.
2: This was due to an error in assembly of the booster, where the cables to the Little Joe II's pitch and roll actuators were inadvertently transposed; thus, a pitch command would instead cause the booster to roll, and vice versa.