A simple answer to this mess (from my point of view):
Time bombing an existing free version is silly and an easily avoidable PR nightmare. Pull the time bomb, release the code for v6.6 to the community and let that version of the source go. Do not include any VS integration, just the stand-alone part. The community has some smart folks who may pick up the ball and run with it. If not, then the fault is laid at the feet of the community for not developing the tool any further. I don't know what kind of agreement Red Gate made with Lutz Roeder when they bought the code from him, if there is anything there that would prevent you from releasing the code to the community. If there is, please speak up and that will squash that argument. If RG can't release the code, then at least pull the time bomb and make v6.6 the last free version.
I find nothing wrong with charging for v7 and beyond. Improve the decompilation engine, UI, integration with VS2010+, etc. A new version is a good breaking point if you have to do something like this. $35 is pretty reasonable considering most console games cost nearly double that and you may only get a few days/weeks of use out of them (depending on the game). There are many programs out there, like Reflector that went from hobby/free to a paid product with support, TestDriven.NET and NCover come to mind. I recall TestDriven.NET going through this same mess a few years ago, though they didn't time bomb the free version, just gave it a lesser feature set and very restrictive licensing.
No well-run business could afford to put full-time people on a product that brings in no money, which is probably why development on Reflector has slowed in the past couple of years. Tough to justify programmer time on something that has little benefit to the company. Microsoft can do it because they have lots of resources and want to sway developers to their tools. Get them hooked and make them want to upgrade. It's harder for Reflector to do that, most people want it to decompile code, once they get that, they are done. The only thing that would sway them is a better decompilation feature set (like inner classes, better property generation, etc). Free version can't do it, Pro version can.
Red Gate hit the same wall as many other companies who try to sell product in the Linux community: you can't sell to folks who only want free. A good lesson learned and it is often learned the hard way. The advantage you have is that decompilation is not easy. If it were there would be lots of Reflector-clones out there which there aren't.
So, make a clean break at v7. Remove the v6.6 time bomb and release the code to the community, if possible or at least just remove the time bomb. Let the community figure out what to do with it. Charge for v7 and make it the product you want to. Those who are interested will pay for it, those that aren't wouldn't purchase it anyway.
Just my $0.02.