As a practical matter, no difference.
But if I had two otherwise identical samples I'd probably go with the high resolution version (human nature).
Back to the more practical side -
The guys who do
blind ABX tests will say you can't hear a difference between a high-resolution original and a copy downsampled to "CD quality".
And assuming you are mixing,
mixing actually increases resolution. Mixing is done by summation so if you mix two full-scale (0dB) 16-bit tracks it takes 17 bits to hold the summed data. Or, since you usually lower the volume before/after mixing you can get more resolution on the "quiet end".