In common with other methods that remove the water signal at the start of the pulse sequence (e.g. presaturation) there will be a bleaching of e.g. CaH resonances but there is no transfer of saturation hence the exchanging NH resonances are not attenuated or lost from the spectrum.
Relatively good water suppression is obtained for DQF-COSY because the mixing time is very short. For experiments with a longer mixing time, such as NOESY, the water signal would start to recover leading to inferior suppression.
1) Piotto, M. Saudek, V, and Sklenar, V., J. Biomol. NMR., 2, 661 (1992)
2) Rance, M., et al, Biochem. Biophys. Res. Commun., 117, 479 (1983)
3) Altieri, A.S., and Byrd, R.A., J. Magn. Reson. B, 107, 260 (1995)