... except it looks absolutely awful.
Despite what I thought was adequate bypass capacitance, I've got absolutely enormous switching spikes (600 mV p-p, from -2.7 to -3.3V on a nominally -3.0V rail).
But hey, it's turning on so at least I have something to go on for debugging.