As I pointed out earlier, there is no reason why "the consensus of mining nodes" would build upon a crazy large block, which implies an implicit block size limit every mining node, individually, decides for himself.  We are forgetting here the consensus role and the voting power that mining nodes have: they DECIDE on which block to mine.  They DECIDE what the consensus will be.  As such, if there is block 50 005 which is still normal, and someone propagates a 5 GB block on top of 50 005 as candidate 50 006 block, it is upon the rest of the mining nodes to decide whether or not they will orphan this block, and make their own 50 006 block, or whether they will mine on top of that 5 GB block.  If a consensus of miners decides to mine on top of that 5 GB block, then it means that bitcoin's consensus is that such a block is "good".  If they decide to orphan it, then it means that even if that 5 GB block is "legal", miners don't like it (essentially because it is spam).  Moreover, those mining on smaller blocks have a network burden advantage.  By the time you've downloaded that 5 GB block, you've lost a lot of precious hashing time - unless the network is so swift that 5 GB blocks are not an issue.  

So you can think that every mining node has his own "limit of block size on top of which he will not mine", for reasons of principle, for reasons of network efficiency, and for reasons of cost.  Of course, if he sets that limit too low, he will never mine.  If he sets it higher than his peers, no problem.  There will hence be a kind of unspoken "market size" of blocks that miners in general accept.  That automatic self-control mechanism would in any case have prevented that the blockchain would be entirely stupidly be filled with GB of nonsense per day.

There was strictly no reason to put in a hard limit.  Most probably, Satoshi thought that it would be a formalization of this "unspoken limit" on which miners would decide.  Maybe Satoshi only saw this as a "mining strategy" as he was also writing the sole mining software out there.

The real long term problem bitcoin was facing is that no block limit would lead to very low fees.  Very low fees would start to be a problem when bitcoin's issuance would start to dwindle, and wouldn't pay enough for the mining security. The finite amount of bitcoin issuance, which is its main publicity and selling point, is in fact a huge monetary design mistake which makes bitcoin unsuited as a currency ; but on top of that, it forces a delicate fee market that has the potential to kill the usability of the system as a whole.

But the finite block size never played a useful role, and its use as a "spam protection" is obviously wrong.  It has only negative effects.