I didn't necessarily mean short the LED, just remove it to eliminate optical crosstalk. But you are right, shorting it would keep the same current, which could possibly induce small currents in nearby wires, so was worth a check.
If you have the emitter in the same housing as the receiver, then yes you will require double the range (20 foot, gap - 40 foot for there and back) but that still isn't unheard of with standard IR receivers and emitters.
I don't see why you would need a mirror, I would have thought just a white strip or a retro-reflector (the kind used on bicycles, road cats eyes etc..) would be much more appropriate and wouldn't require careful alignment.
I believe you can reduce the duty cycle of the 38kHz carrier and raise the current, but remember - devices used in remotes are "low duty" by use as well, never designed to constantly transmit for hours on end. I wouldn't go over 100mA peak (although many datasheets say you can go much higher). Also, you can double modulate, if you don't mind a slight delay between beam, breaking and output, or if you don't mind missing very fast objects that break the beam for milliseconds. This would require two 555's but will save power, and with 38kHz bursts every say 10ms, plenty of time for the emitter die to cool down so you could increase the peak current. If you only wish to detect relatively slow objects, or large ones where the beam break will last longer, you can of course reduce the burst frequency further.
Detecting would require a missing pulse detector - starts to get more complicated.
One thing that will increase range, and most likely improve the optical crosstalk situation would be decent optics. I'm not suggesting lab grade IR lenses here (we're still talking light here, not long wave IR) but a couple of cheap convex lens, and a couple of plastic tubes to mounting them in. With the lens for the emitter to make it diverge less, and the one for the receiver to make it 'see' less.