This is absolutely true. Some keynote speakers just speak rather slowly, or repeat themselves and you can easily get away with 1.3x speed. Even though it sounds funny.

I really wish there would be some sort of plugin that could select an even higher speed for silent/quiet parts.

I figured "how hard could it be" (having some experience writing tiny softsynths in the past I wouldn't even mind writing a low quality pitch-preserving resample filter) but digging into it, coding a plugin for VLC is pretty arcane. At least, I couldn't figure it out without having to dig into all sorts of audio/video decoding issues I didn't really want to get in to.

If anyone has any tips it would be greatly appreciated. I cannot promise I'll manage to code that particular filter but at least I can have some fun writing crazy VLC filter effects :)