> maybe the stress test was measuring something different?

The Fed is measuring the loss on bank loans to the private-credit lenders. A 10% portfolio loss shouldn't result in those lenders defaulting to their banks.

By my rough estimate, one can halve the portfolio loss rate to get the NBFI-to-bank loss rate. So a 10% portfolio loss means we're around a 5% expected long-run loss to the banks. Which is still weirdly high, so I feel like I must be missing something...