a setup like this should work.
arrays are pretty simple, simply put its a list. so if you have a array of integers its just a list of numbers. lets do a little exercise as an example, open up a spreadsheet program (excel, openoffice calc, etc). now fill in column A with a series of numbers. if you notice the row numbers to the left of column A those would be the array index and the information in column A would be the array data. now lets say you have 10 entries in the array or 10 rows worth of data in the spreadsheet. if you delete a row then the table will shift to compensate. so if you delete row 5 then the data that existed in row 6 would shift and now occupy row 5. this is the same as how the remove index node works. so this also tells us that the length of the array is dynamic.
if you wanted to reuse the array for another loop further down the graph you could always just have another set node after the loop complete pin. this would in essence repopulate the array.
Im also half asleep right not so it may not be 100% right.