Further to BJJ's explanation:
Think of it as your insurance policy. As highlighted, some aircraft are not guaranteed to achieve minimum climb gradients with an engine failed until, for example, gear and flaps are retracted and the failed engine is feathered. You can derive this from the performance graphs for the aircraft type.
If the aircraft is not able to climb above obstacles, you are committed to landing, and you must have sufficient visual reference to achieve this. It would be a risky manouvre to depart into a 50ft cloudbase, lose an engine and be unable to climb, then be forced to try to land with an RVR of 200m!
Thus, the height at which the aircraft can deal with the failure and achieve an obstacle-clearing climb gradient becomes the number you use to set a minimum RVR for take off.
An example:
One of the aircraft I fly has a worse climb performance with flaps extended than with them retracted, and the flaps are not always needed for take-off. So in the case of an engine failure it can climb away without flaps extended from a lower height than it can with them extended. Therefore, the minimum flapless RVR is 300m, but with flaps it is 500m.
Another aircraft made by the same manufacturer but with one less engine can achieve the same climb gradient irrespective of flaps extended or retracted. Therefore the RVR for that type is 300m.
As for when you use these minima, the answer is whenever you take off. They should be written into your Ops Manual and you lay yourself open to investigation by the CAA if someone reports you departing with an RVR lower than the minimum allowed.
All of this is of course under public transport rules. If you are non public-transport, then the aerodrome minima apply (these are usually lower), but the risk for you remains the same if your donkey quits at 50ft...