A probe's frequency rolloff is much too poor and poorly defined to be of any use as an anti-aliasing filter.
yes, probe frequency rolloff is not the best, but it helps to reduce high frequency components to eliminate aliasing issue. Especially for a slow time/div modes which is required to analyze low frequency audio signals. And it's better than no filtering at all when you use very wide bandwidth probe with low bandwidth oscilloscope.
The scope should incorporate the specific anti-aliasing filter that the scope needs. If not, then get another scope. (There are rare counter cases, but they are not relevant to a beginner)
Yes, good oscilloscope should have antialias filter on the input. But using narrow bandwidth probe helps to reject unwanted high frequency components much better.
For example oscilloscope antialiasing filter may reject some component for 30 dB. And narrow bandwidth probe provide additional 30 dB rejection. In total we have 30+30 = 60 dB rejection. When you use wide bandwidth probe you will get just 30 dB rejection, which is not enough to not see unwanted aliasing.
Cheap oscilloscopes, like these 70 MHz may not have antialiasing filter at all.
For example FNIRSI-1013D shows aliasing distortions because it's specification bandwidth is higher than the half of it's ADC sample rate.
But the thing going to be much worse if you needs
to measure low frequency signals (this is exactly the case of topic starter) and your oscilloscope has too small memory or too slow DSP (which is guaranteed for a cheap oscilloscopes). In such case oscilloscope needs to force reduce ADC sample rate for slow horizontal resolutions. Because it will be unable to capture and process enough sample count at full ADC speed for a long time interval (which is required to see a low frequency waveforms).
In such case internal antialiasing filter of oscilloscope cannot help, because it's cut-off is much higher than half sample rate for selected mode and your waveform will suffers distortions due to aliasing. This is a standard issue for a slow time/div modes in a cheap oscilloscopes.
Probe bandwidth doesn't matter. It may be 50 MHz or 500 MHz. It only must be calibrated with the exact input. That's all. Wider bandwidth is almost always better then narrow.
this is false information. Using expensive wide bandwidth probe with a cheap narrow bandwidth oscilloscope may leads to a waveform distortion due to aliasing. How much worse depends on the used antialiasing filter and osclilloscope bandwidth to oscilloscope sample rate relation and also oscilloscope memory size and it's DSP/CPU speed. For a cheap oscilloscope it may be very bad and very noticeable.