So those are not single-channel test gateways? If they’re full gateways, then still their range might be shorter (from some direction) simply due to their position. Like: surrounding buildings, antenna height, and when they’re installed inside, maybe even behind insulating glass.
A proper node should perform frequency hopping: it should randomly use a different frequency for each uplink (at least in EU868). But it will only change its SF to improve its reach.
Single-channel test gateways often listen to a single frequency, on a single SF. Like Kersing wrote:
Some single-channel test gateways listen on multiple SFs for their single frequency, but then their reception range is decreased due to using CAD.
Full gateways are ready to receive on all frequencies and SFs of a given frequency plan.
Not automatically. You could implement sending confirmed uplinks and keep track of which frequency and SF works. But that feels like a huge waste of time to be able to use test gateways, especially as the number of gateways could change any time. Also, you’re limited to 10 confirmed uplinks per day.
For testing, if you know some gateway’s settings, see How to make RN2483 node use a fixed channel and spreading factor?