Well, client-side validation is mathematically secure, while SPV is economically secure.

I.e. it is secure if you make several assumptions about economics of the whole thing.

In my opinion the former is transfinitely more secure than the later.

But it's more of a philosophical question, sure.

The good thing about PoW-based consensus is that it is robust against version inconsistencies and various accidents of this nature up to a certain degree. But you hardly can depend on that: 

You know, The Great Fork of 2013 was resolved through human intervention, Bitcoin nodes were not smart enough to detect that something is going awry on their own.

Naive proof-of-publication is very fragile in that respect, but you can easily bring back robustness.