We have seen this 'longest chain' verbiage where 'valid' is notably excluded for some time.  I take it as a backup or auxiliary attack strategy.  Every opportunity is taken to lay the foundation participant misunderstanding however.  That misinformation effort seems to have picked up again over this last month here on bitcointalk.org.

Basically, if a bloatblock (for instance) can be inserted just one time, it would fork the chain and obsolete Bitcoin core software.  A combination of political/enforcement activities against mining centers in 'aligned' jurisdictions and blocking of those operating in non-aligned jurisdictions would do the trick.  We've already seen Hearndresen work the political end of things in surreptitious meetings and pressures on large mining installations and their management structures.  Thankfully these activities appear, outwardly at least, to have been relatively unsuccessful and it seems to have spurred an uptick in alternate related attack strategies.

A nicety of this attack would be that operators could shed crocodile tears about the 'unfortunate' situation, but say "longest chain, what can we do?"

The workability of this attack strategy relies on several things:

 - There is widespread misunderstanding of the 'valid' part of 'longest valid chain.'

 - Bitcoin core is molded into a situation where it fails if it does not act as a near real-time system.  That is to say, those who rely on it are not able to take a 'hiatus' for the amount of time it takes to mount and win a counter strategy (which would be more political than anything else.)

Subordinate chains could continue to operate during a hiatus in core function, but in degraded form.  This 'degraded but still functional' aspect would take much of the pressure off the core (aka, the 'backing') solution and allow it to outlive an attack such as the described.  Such an attack would be expensive and difficult as a function directly related to the time over which it must be sustained.