I think you could have avoided a pain point here by use of A/B testing: show premium users the same page you've always show them. Show some percentage of free users the new treatment ("premium, but for free") and the remainder the old default ("like premium, but less of everything you care about"). Then measure conversions. If your "premium, but for free" users don't buy statistically significantly more domains (enough to offset the premium revenue), then you can abandon the test and your premium users never have to know you had a crisis of confidence in your business model.

Its easy to say. Its hard to do, particularly with dynamic sites. Google Website Optimizer is a halfway decent solution if you can get it to work. I hope to have another halfway decent solution available for Rails users sometime in the not-too-distant future, as I'm sort of fed up with the GWO "all tests incur 1.5~2 seconds of extra load time as we figure out what URL to redirect to, because that totally won't murder your conversions" problem.