Be aware the maximum lookback stated in the documentation is "31 days", whether this is simply 86400*31 seconds or some other period isn't stated. But if you miss a run for some reason, or there is a network/API/other glitch (these things happen), you may run out of time to recover before losing data.
I recommend use the Python library with async IO to run on all networks concurrently (underlying is rated limited, but the library handles that for you) as you will be issuing call-per-network, this will run faster than network at a time.
Fwiw for testing I currently run it daily with 600 second resolution and build a long-term time series, same as the approach for a number of other things we measure, the automation handles some re-runs itself, or we can recover manually by setting t0-t1 as appropriate.
