Simulate out-of-disk situations in chutney and/or unit tests?

Tor has a bunch of weird bugs when it runs out of disk space.

One of the crazy ones that has been driving us mad for more than a decade is that relays somehow decide to start publishing a new descriptor once per second, forever.

Can we make a version of the unit tests where all of the disk interactions act like the disk is full?

Can we make a version of the chutney network tests where the disk pretends to be full?

This idea seems like a great one to add to our testing toolbelt.