What I mean is that it allows temporary support if it can do so without danger of causing insolvency. If the choice comes to down to not giving support or becoming insolvent, it should opt for not giving support.
A “full” network isn’t immediately a critical failure. The network first has to shrink before data loss may occur. In addition, last time I checked we have two primary copies, two backup copies and two sacrificial copies of every chunk. The network is “full” when all the sacrificial copies are gone. In the case of the network shrinking after it’s “full”, vaults can (and should) opt to first delete their backup copies to make room for any incoming primary copies due to churn. Sure, this is an undesirable situation, but the network may very well survive such a scenario without any loss of data.
In contrast, insolvency (hitting the supply cap) is I believe far more critical, since FR can become 1 while still not resulting in any issued coin. If the network relies on economically minded farmers to any significant degree, a lot of them may go offline. If this then leads to a “full” network, I would be extremely surprised if the network survived that. Both algorithms would be dead.
This could theoretically be done as well by flat out refusing new uploads at a certain level. I admit that a gradual discouragement is more desirable (continuity of a service using SAFE may be worth fortunes), but at some point there’s no practical difference between a horrifically high store cost and a flat out refusal of accepting new uploads.