lensing of the CMB

Could one imagine a situation where all structures in the CMB are wiped out by lensing?

Typical deflection angles in CMB-lensing are much less than the typical size of structures in the CMB, so the hot and cold patches appear weakly distorted. One would need deflection angles comparable to the size of structures for wiping them out. This can be seen in the CMB-lensing formalism, where the variance of the deflection angle appears in a convolution relation operating on the spectrum.