"Steady state" means the average amount of data sitting on disk once the policy has been running long enough that old generations are being deleted at the same rate new ones are being created — the system has reached equilibrium rather than still filling up.

For example: if a project rolls over every 1 day and the policy deletes data after 10 days, at steady state you'll always have roughly 10 generations sitting around (1 rolling out the bottom for every 1 added at the top). That's different from a "right now" snapshot, which might show fewer or more indices depending on when you happen to look — e.g. right after you manually deleted some, or right before a big batch is about to age out.

I used it to distinguish two different numbers:

• "Right now" — whatever's actually sitting there at this exact moment (e.g. nginx showing just 1 cold index because you manually cleaned up Thursday/Friday)
• "Steady state" — what the typical ongoing footprint will be once the policy has been running consistently for a while, ignoring temporary blips like manual deletions or a recent burst of writes

It matters because if I only quoted you "right now" numbers, a one-off manual cleanup would make a policy look better than it actually behaves day-to-day. Steady state is the more honest number for "how much disk will this policy normally consume."