I built a reactor that detects when power is being drawn by a MFSU and activates the reactor to recharge it, and when that draw stops, turns the reactor off. When it is fully shut down there is only one 3-second timer and a 0.2 second sequencer (which AFAIK can't be turned off like timers in RP). The sequencer is used in the bucket flow failsafe because I was told it is possible for redpower timers to fail serverwide, but not sequencers. The reactor watches the bucket rate and if it is not well within the parameters it shuts down the reactor until a human cancels the shutdown; so far this has never happened since its completion. It uses dual bucket filling and cycling mechanisms, and uses only RP2 and IC2 parts. It produces 488+EU/t at the moment, it can produce more but I chose to use the entire two top rows for buckets due to its unmanned nature and the rest is uranium cells.
When it is activated, it initializes the bucket system a few seconds before it activates the reactor; on deactivation it shuts the reactor and the bucket flow to it down but leaves the components capturing rejected buckets and pumping them back to the bucket storage chests active for an additional 45 seconds.
This kind of setup may not be as useful for primary bases or manufacturing plants or the like, but it was very appropriate for my purposes, which was as a teleportation hub. Generation of power most of the time was a waste of uranium; when a teleport actually occurred, however, recharging needed to be as fast as possible.
The power draw detector is the device with the water mill.
The draw detector, MFSU, and the startup circuits are not within the reactor chunk but if they unload should not cause any reactor failure. Worst case, the reactor would run indefinitely, until its uranium depleted or the control system chunks loaded.
Its a bit disorganized, but that is because I sort of made it up as I went.