The incentives are what they are. Sure some miner can play the role of 'attacker', and include a transaction that takes an inordinate amount of time to validate. My claim is that incentives are aligned to render this a non-problem.
Riddle me this: built off a parent of the same block height, a miner is presented -- at roughly the same time:
1) an aberrant block that takes an inordinate amount of time to verify but is otherwise valid;
2) a 'normal' valid block that does not take an inordinate amount of time to verify; and
3) an invalid block.
Which of these three do you suppose that miner will choose to build the next round atop?
Riddle me this: built off a parent of the same block height, a miner is presented -- at roughly the same time:
1) an aberrant block that takes an inordinate amount of time to verify but is otherwise valid;
2) a 'normal' valid block that does not take an inordinate amount of time to verify; and
3) an invalid block.
Which of these three do you suppose that miner will choose to build the next round atop?
Your question is poorly constructed because you do not define what qualifies as an "inordinate amount of time to verify."
As you know, the problem with troll blocks is precisely the fact they are valid, despite being trollish.
Who said anything about invalid blocks? What a red herring. Miners running Bitcoin Core compatible nodes don't mine invalid blocks, although there's no accounting for what zany harebrained scheme the lads at Unlimite will come up with this month.
Empty blocks take no time to verify, yet miners only create them occasionally.
Miners' decisions depend on the amount of fees in the blocks, their software/hardware/network configuration/capabilities/limitations, what they expect other miners to do and/or their strategy for attacking other miners (game theory), etc. For example, a computationally hard block construed in order to clean up a bunch of dust was intentionally mined not very long ago. That functionality would be lost if we used Gavin's artificial 100k/tx limit.
Any more silly questions?
Are you now ready to join us here in the real world, where O(n^2) attacks are a problem, or are you going to stick with the wishful thinking, hand-waving, and "Because Mining Incentives!" slogan?