Hi Alex,
From what I understand, this value was chosen several years ago during the model tuning stage for either a European domain or African domain simulation (I can't recall which). I'm not aware of any reference regarding this choice. I think this choice is justifiable if one assumes that most clouds are smaller than a model grid cell.
However, a relatively recent paper (Wood and Field, 2011; J. Clim) shows that the number of clouds larger than ~200 km size is approximately equal to the number of clouds smaller than ~200 km size, so I wonder if this parameter choice (and associated small-cloud assumption) should be revisited.
Kind Regards,
-Travis-