My Lab contains equipment good to about 1.5GHz. Despite that, I prefer to use a good quality dual trace 20MHz scope. The number of times I've needed to go beyond that in the last four years I can count on the fingers of one hand - and then it was only because the faster scopes had the ability to display on-screen wave analysis statistics rather than the need to actually display a fast waveform.

Ultimately, it's all down to what kind of applications you're playing with. I know I'd rather have a scope that is in-calibration and I can rely on to tell me a resemblance of truth, than one whose results can't be trusted. There's more to chosing a scope than raw MHz on the timebase.