Why is the check < 12? That looks like the opposite of what I would expect.

The data change every hour, but is available 12-36 hours ahead. So when we have less then 12 hours of future data available new data should be available. We do not know exactly when the new is available, so we might need to retry.

Why shouldn't we let the scan interval decide on the update frequency?

The data change every hour, but we want to retry if no data are available.

If the data hasn't changed since last update, making a state update will not write a new state to the state machine. So you could just try to do that without making the check here. Are there any downsides, you think? It would make the code cleaner.

I am not sure I understood your last comment.
But if I do not return here, the _update_current_price will be called every minute

Yes, and if all values including state and state attributes remain the same a call to do a state update, will not write a new state to the state machine, so will not clog the database.

The only thing that seems to change, while fetched data is the same, is the time_diff variable that is computed from the current time of the call. What will happen there?

Yes, there will be no state change and no new state to the state machine.
But I though it would be good avoid calling _update_current_price each minute, and then also avoid iterate over all prices each minute.
So the return here is only to avoid unnecessary cpu load.

👌