Position sensors are probably fairly cheap (worth trying regardless). I replaced the cam sensor a couple of years ago. Symptoms were different then and it gave me an error message.
I bought a fuel pressure gauge (in anticipation of that part of your answer) and I will try that today. Presumably, this could also point a finger at the fuel pump? I wondered about the pressure regulator (FPR, right?) too, but it didn't seem like something that could fail in this way. I also bought a little hand-held vacuum pump that should provide the definitive answer on that, right?
Air flow sensor was pretty high on my list (because of the easy restarts), but looks pricey. I have one other data point on that - I can make the car stall from idle by tapping on the top of the airflow sensor. I don't know if this is normal or indicative of a problem because I only have one subaru to try it on.
Fuel pressure is in spec at idle with the FPR connected (29-30) and disconnected (38-40). When I open the throttle, the pressure jumps to the disconnected value. I assume that this is what it is supposed to do (i.e. its function). I pulled vacuum on the FPR and it held (with engine running so that there was fuel pressure on the other side). The fuel pressure seems to hold when the engine shuts off so it is not bleeding off anywhere quickly. When I stalled the engine by tapping on the air-flow sensor, the fuel pressure did not seem to drop at all, so whatever effect this has does not seem to be fuel starvation.
There were 4 wires on the air-flow sensor. All were dirty, but they looked like white, black, red/yellow, and yellow. White (my left side looking in) was the only one where I saw voltage. It showed 0.3V with key on and engine off, 1V at idle, and it jumped up to around two volts when I opened the throttle plates. I assume that this was it sensing a greater air flow.
The date of manufacture of my car (according to the tag on the door frame) is 03/96, so it is a little bit newer than the service bulletin regarding flakey crank/cam angle sensors. I ordered a copy of that online from NHTSA. I will go ahead and buy those anyway just in case.
A little more experimentation indicated that tapping around the electrical connector on the air-flow sensor was the way to make it stall (rather than tapping elsewhere on its body). I tried disconnecting this, which I have never done before. It looked fine inside (no burnt looking pins). After I reconnected it, I could no longer reproduce the stall by tapping on it. Could the whole thing be as simple as a loose/dirty electrical connector?