I meant best in a way that included the network effect, not in a moral or technical sense. Substitute 'most effective' there. On a side note, I think you're not realizing that a 51% attack like you describe would be a simple denial of service from all of the current software clients perspective. The user experience WOULD be disrupted and users would have to choose what to do if anything. The no-op choice would be have unspendable coins.
Not necessarily. Ultimately, the users just have to attach larger fees to their txns. This won't be unusual at all. Remember the plan is to jack up fees progressively as the block reward falls. It is not a disruptive change if the fee increases somewhat more rapidly than the initial plan. A wise monopolist would introduce his fees gradually. He might even allow no-fee txns at first and just increase their delay in entering blocks over time. There is no need for any discontinuity from the end users' perspective.
Anyways, if there was still block reward you can just start by not screwing with fees at all and simply expropriating other miners. This does not affect users in any way, but could be highly profitable.
Compare these changes to finding a new client to download and starting to use coins that are not accepted anywhere.
Finally, you are still ignoring that attacks are not harmful to users under PoS. This perhaps even more important than the fact that PoS makes attacks more expensive to pull off. (Of course the two facts are related)