Answer
Disk sizing for a REF environment is driven almost entirely by actual data volume rather than user count — if the environment stores large binary/media objects (documents, attachments, images), the tablespace holding those LOB objects will dominate the sizing far more than transactional table growth would. If REF is going to be periodically refreshed via a clone from Production, plan for REF's disk to end up needing roughly the same size as Production's actual database footprint at clone time, since a clone carries over Production's full data volume rather than a scaled-down subset — this is the most reliable practical basis for sizing REF's disk, more so than any fixed vCPU/memory-based rule of thumb.