I don't quite understand the meaning of loopback. Do you mean a long fiber and then use a repeater to drive back to the beginning of the loop to keep the data circling forever?
If so, you calculate the velocity of the fiber, then find the delay from the beginning to the end of the line. Then use the data rate ( bit/sec) and multiply by the total time delay to get the amount of storage.
Do you mean somehow you inject data into a big loop and let the data propagate by itself forever? That will not work as it can travel either direction and you start creating standing wave from both direction!!!( nobody can tell which direction the data propagate!!! You are not there to direct the traffic!!!).
Let me take a wild guess, you are hopping to eliminate all the active circuit( repeaters/ multiplexers) that has a data rate limitation, using a mono mode LOSSLESS fiber so you can pump very very high speed date rate. In this way, you can store the maximum amount of data in shorter line loop.
Actually I thought about it a little more, if you use repeater and multiplex and a long fiber line, it is like dynamic RAM that you reflesh with the repeater, then you can read or write new data through the mux. You keep track of the bit in the loop with counter which serve as address lines in DRAM.
Problem is the speed in fiber might vary slightly, it is not exactly like charge in descrete cell. Also the data storage is limited and depends on the fastest repeater and mux. Say using OC192 10Gbs. Say each data occupied 1" on the line, you need 1000' to store 12000 bits of data. Say if you use very advance modulation like QAM, you might say improving 16 times, still that is a lot of hardware for very little data storage. Might be cheaper to do storage in static RAM of very wide data so you get the equivalent speed up.