After some research (and a bit of a headache), I came across some interesting points related to what I was looking for in the post below. I also have some suggestions and would like to check if they make sense and if they could potentially be implemented in the future.
Currently, the only way to update a collection is by refetching it entirely. If you need a partial update, you must store the updated data in a variable. - (is it still the same?)
My suggestion if makes sense > Additionally, collections automatically generate a mirror variable, which can be used as a state variable. This eliminates the need to create one separately and serves as the ideal place to store updated data, as mentioned earlier.
A valuable enhancement would be the ability to clear a collection through a workflow.
Another key limitation is the inability to perform partial updates on collections, which would be highly beneficial.
My suggestion if makes sense > The ability to fetch only the necessary data using filters in workflows would greatly improve efficiency.