@3point: In my experience, that works well in theory or in ideal situations (e.g. 50 fps–>25fps). In all other situations it’s going to be a big royal mess!
My students are expected to make short videos using all sorts of input materials they could get their hands on. This means using videos from common videosharing sites such youtube/vimeo as well as from other sources (e.g. cell phones cameras, web cameras etc). Trying to figure out how to combine the various frame rates (ranging from 10 fps cell phone cameras to 60 fps or more) is really asking for trouble especially when the audio has to be synced as well. Each file has to be treated separately before it can be imported into a single fps VSE project for editing.
Things are looking brighter when the audio tracks are not needed or no sincing is required but you can never bet on it due to Murphy’s law.
@Gwenouille: yes, in principle, it can be done but be prepared to fiddle with it. In most cases this requires some experimentation to get it to work. I would advise against it really, esp if you’re on a deadline.
Alternatively, I would strongly recommend using Avidemux to pre-process sources before using the VSE for editing and rendering. It has an excellent resample fps filter and if you render out to a lossless codec then there is no image quality degradation. I have found it to be a great compliment to Blender’s VSE esp for fps tasks. Not to mention, of course, the fact that you can very easily trim a video and save the file without the need to re-encode it!