"overwhelmingly" only if you judge mainly by print sources.  In my experience, manuscript scribes from Nepal and from Kashmir routinely use sibilant-assimilation sandhis of the form namaśsivāya.   I think this is also true of MSS from Kerala and probably elsewhere too.  I think the dominance of -aḥ before sibilants is an artefact of normalizations in nineteenth and twentieth century editing and printing practice.  Like the occlusion of extremely common MS forms such as karmma, dharmma, etc.  Word-final nasals too are normally left as savarṇa nasals and not converted to anusvāra in early Nepalese MSS.  In short, sandhi practices are much more flexible and varied in the manuscript record than in printed texts (while remaining Pāṇinian). 

Best,

Dominik