In other words, this is a direct credit risk to the bank, not the more indirect reputational or relationship risk of potentially unwittingly or negligently facilitating a scam described in TFA.
From what I can tell banks generally don't rely on airtight logical guarantees. Rather they have some kind of exposure on every transaction, which they work to reduce. This isn't the most efficient system, but they deal with more types of fraud than just fake checks.
In other words, "at the time we paid/did not reverse this, we thought the payer account was good for it, but as it turns out it was funded via transactions that ended up being reversed, and now the account is in the red" is not a valid reason to claw back money.
And that makes intuitive sense as well: Only the paying bank has a complete picture of all account activities (including out-of-character/potentially suspicious ones), so it makes sense to primarily hold them accountable.