Just to clarify - here's a chart showing typical BW curves for a 500MHz Nyquist frequency - in other words, the Nyquist frequency of the DS2000 when you have 2 channels ON at max. sampling rate (1GSa/s). Even though most people agree that sin(x)/x should have a sampling rate 2.5x the highest frequency, let's just assume, for simplicity sake, that it can work adequately with 2x. But any frequencies HIGHER than that 2x can cause problems for the interpolation.
The chart shows the hypothetical "brickwall" filter which would be ideal to have (the red line), a maximally flat response curve (the blue line), and a typical Gaussian response curve (the green line). The gray diagonal lines show the area of trouble: where higher-frequency signals can leak through the filter to cause mistakes in the interpolation. With the given response curves, that area is all below -8dB.
On top of this, I've overlaid the results (the yellow line) that someone posted here for the response curve of a 300MHZ "enabled" DS2000 HW v.2. If these results are correct, the area of filter leakage (the orange diagonal lines) increases to a point above -6dB when using two channels.
I would suggest, for anyone enabling 300MHz on ANY DS2000 - that wants to be certain of signal fidelity (unless/until we see further tests), to use it only with 1 channel @ 2GSa/s (otherwise use the channel BW filter).