Threats are inversely related to what's at risk. When you mention extensive key reuse, that can mean a variety of things. Are you referring to the number of notes being published or the number of clients using the same key? I suspect it's a combination of both.

That said, it's difficult to provide a global risk assessment. If someone like ReplyGuy has their key compromised, the impact is minimal. However, if a more prominent figure like Jack, Gigi, or ODELL gets compromised, the damage is more significant—it sucks, but it's still localized to them.

This brings us to a kind of structure or formula: the longer a key is in use, generating valuable and trusted information, the more reputation it builds and the greater the associated risk. There's a (time x value) equation tied to each pubkey, and this value doesn't begin at the network level, it originates bottom up from the consumer level. Ultimately, it's the recipients of the key's information who determine its value across time—the creator doesn't get to decide. 

Think about how nuanced that is, the biggest risk in losing a key is to the person (or machine) that seeks trust from others. The flip side in the risk equation is when that entity breaks the bonds of trust - then every recipient in their network suffers. You don't need to lose your key to do that though.