I think they mean the scenario where the attacker sneaks in and adds a persistent backdoor / sets up privileged accounts for themselves to use / generally makes themselves comfortable, and then sits on their thumbs for a long period of time before activating the ransomware. If the victim restores from the previous day / week / months backup, the attacker can then sneak right back in and re-exploit whatever vulnerability.

The proposition then becomes "Our last non-compromised (as far as we know) backup is half a year old, can we afford to lose months of progress, or should we just pay the ransom?".

Or perhaps the attacker doesn't have a persistent foothold, but can make a credible claim that they do - do you take the risk? Keeping in mind that every moment you burn investigating and trying to discover the extent of the compromise is another moment you're down and burning money.

You might have a good backup from say a week ago which is great to recover from but then you still have to sift out what may have been compromised transaction-wise if the attackers were clever in hiding their attack