LAG are aggregated interfaces and they can indeed be used to prevent (some) layer 2 loops. LAG as in Link Aggregation Group)
Using 2 non-LAG interfaces between the same 2 devices creates a loop.
In the case of a loop, if you’re running spanning tree, one of these interface will be blocking instead of forwarding, preventing the loop, but also percentile the use of this interface until the topology changes (ie: the current one goes down).
If you’re not running spanning tree for some reason, then both interface will chug along, oblivious to the fact that there’s a loop and broadcast packets will indeed keep being flooded on one and received on the other, again flooded, etc. creating a broadcast storm and impacting performance of the whole layer 2 domain and possibly even crashing devices.
A LAG more or less means the interfaces in the group behave as one big (aggregated) interface.
LAG also means you can push traffic on both interfaces for more bandwidth.