>1)  What could be used to economically switch power to the relay?

I suspect that the relay will stay in one state for long periods of time.
If that's the case, you should look into latching relays.  I've used them
in cable TV equipment quite successfully.  Then determine the amount of
energy needed to change state, and the minimum state state expected.  I
would suspect you have some type of hysteresis and debounce, so 3/sec is
probably NOT the minimum duration.  Say 5 seconds.  Find a cap which will
charge in 5 seconds (assume cap is at least 20% larger than it really is
for charging purposes, and at least 20% lower for energy storage) through
your resistor that you pick, but has the energy to flip the relay.  Also
assume the relay is going to need twice as much power as spec'd.

You'd be surprised how little work it takes to make this work out right.

>2)  Is 3 volts going to work as the supply voltage (especially since the
>voltage will decline as the battery is discharged)  Would a higher voltage
be better?

Pick a cell chemistry which does NOT drop off linearly, such as a
nickel-based technology.

>3)  Is there a clever way to reduce the on-time of the LED?
>If I had a faster sensor (in the visible range), I could shorten the
>LED on time.  Is there a low power source-sensor pair for visible
>light?  Clever ideas on reducing the power required for sensing?

Try using the same type/color of LED as both an emitter and a detector.

>4)   Any cool suggestions on power management?

Definitely do some.

Remember battery output changes drastically with age and temperature.

Andy

Say, Gus, I'm on my own again.  SRI is looking for somebody.
   \-----------------/
    \     /---\     /
     \    |   |    /          Andy Kunz
      \   /---\   /           Montana Design
/---------+   +---------\     http://www.montanadesign.com
| /  |----|___|----|  \ |
\/___|      *      |___\/     Go fast, turn right,
                               and keep the wet side down!