If you open project one computer connected to the Cache Server - it will fill in Cache Server with the import results(that is correct). But when you open the same project on another computer connected to the same Cache Server - it fills in Cache Server again (that is wrong).

How we can reproduce it:
1 I have a GitHub repo with 2.5Gb project
2 I clone that repo to 2 machines - A and B
3 Machine A has Cache Server running
4 Editors on both machines configured to use Cache Server and that works
5 On machine A I open project - it imports and fills in cache folder of Accelerator (everything works so far)
6 (Problem happens on this step) On machine B I open project and it imports assets and tries to fill in Cache Server on machine A again - it sends a lot of data to machine A

Expected results: Editor on machine B should pull import results from Cache Server(Accelerator)

Reproduces on both 2019.3.0b6 and 2020.1.0a9
Logs from Editor 2020.1.0a9 from two machines are attached (look for "Editor_machineA.log" and "Editor_machineB.log")