You can check their demos on every OS you would like to.

Actually it is a very stable technology and fast. The reason is, they are using some proven approaches like pre-rendering light maps from lots of static lights (like WED, Source engine and others). But they also add additional information like some light vectors to illuminate materials very real. This combination of lightmaps, shaders and light direction information makes it fast and good-looking at the same time.

In each indie engine you have to fake it. You can also render a lightmap in second uv-set and overlay it over the tiled textures. But then you have to add shaders and dynamic lights to illuminate shaders. But this approach will look worse and will render slower. You cannot use the same amount of dynamic lights, but you can use tons of static ones for a lightmap. And because of that the lighting direction will be wrong in most cases. Besides that the usage of many dynamic lights in a forward renderer will slow down rendering.

Currently the trend is: If they render every lighting in real-time, then they use deferred rendering. This is not very fast with a single light source, but it is much faster when you use tons of lights and shaders (CryEngine, Torque3D).
The other direction is what UDK and Source are doing: combine static lighting with additional normal mapping and lighting information.

Both solutions make lots of sense.