This is where long memory comes in. You slow the scope down to be able to capture the slowest waveform. Once the waveform is captured, you can zoom into the trigger wheel waveform at any point of the 02 sensor cycle. The bigger the capture memory, the more samples can be captured giving you more high frequency resolution.
For example, with the Rigol DS1052E, if you set the timebase to 200mS/div, and enable long capture memory, then with 2 channels, it is sampling at a bit under 5uS per sample. So that means there will still be about 40 samples per 4800Hz waveform, so that is pretty good. With long memory disabled, the scope would only be sampling at about 3KHz which is useless for the 4800 Hz waveform.
There is also delayed sweep. This is where you trigger off the O2 sensor waveform, but the scope is set to only show, say, a 1mS window. With the delayed sweep, you can move this 1mS windows backwards and forwards in time so you can investigate both waveforms at any part of the O2 cycle. This method doesn't need the long memory, but you can still use it. The delayed sweep will give you the best resolution, but you are only able to see one slice at a time.
Richard.