Can someone explain the difference between using a 'send' effect and an 'insert' effect. I've used both, but don't really get the advantages of choosing one over the other.

Use a send effect if you want to treat a bunch of tracks with the same effect - like Reverb or Delay.
 
Use an insert effect for a specific treatment to a single track - like Compression or EQ
 
Sometime the lines get blurred.
 
This of a Send as a COPY of your source track. This copy is then blended in with the original along with the Effect (nearly always set to 100% wet in the case od Reverb/Delay)
 
Another way to look at it, think of Insert Fx as a SERIES effect, think of using SENDS for parallel processing

In the hardware mixer days, before the digital revolution, one did not have an unlimited number of FX instances like we do in the digital world.  So, an engineer may have only one reverb unit.  Now, if that engineer was live mixing three vocalists and wanted reverb for all of them, he'd/she'd put that reverb on a bus and use a send from each channel to send some of the audio from each vocalist to the reverb unit.  The wet output from the reverb would be presented to a bus which would be mixed in with the overall mix.  The result would be that all of the dry sound from the vocalist's channels would be in in the main mix and the appropriate amount of reverb return was mixed in - all with just one reverb unit.
 
This method was fine for something like a reverb, but if you needed dynamic control of a channel with a compressor, you'd need to put it in series with a channel.  This was a situation where you would insert the compressor into the channel and generally you needed a channel of compression for every vocalist or instrument you wanted to compress.
 
Things have changed a lot since the hardware days - mainly that we have effectively innumerable instances of things like reverbs, so people have taken to just inserting them into the channel they want, rather than setting up a send/return for a reverb.  That said, if you want to configure a compressor to control dynamics (as opposed to parallel compression for effect), you generally need to insert it.  This is because inserting FX places the FX in series with the signal flow, whereas placing an FX into a bus and using a send generally results in the FX being processed in parallel with the original signal.
 
Cheers,
Bill.

Also, there is something to be said for using a single reverb for most tracks, namely that you are putting things in the same "space". Often I'll still have three reverbs in a given project but I would definitely not put separate reverbs on every track.

I use different reverbs for different applications, as suggested by Mike senior in his excellent book.
 
He suggests you use any combination of the following, provided the project calls for it.
 
Size
Blend
Tone
Size
Spread
 
I usually have a reverb for size and one for blend. Sometimes I'll have an extra size 'verb for different effect, combining early reflections of one with the tail/decay of another, as an example.

Ricebug
Can someone explain the difference between using a 'send' effect and an 'insert' effect. I've used both, but don't really get the advantages of choosing one over the other.

I think this guy explains it pretty well IMHO


https://www.youtube.com/watch?v=lhlUZKXLdZE

http://blog.cakewalk.com/effects-placement/
 

An insert effect such as a compressor affects the whole track. A send effect such as a reverb 'sends' some of the signal to the effect bus while leaving the original signal unaffected. The exception would be an effect that has a wet/dry control. That's how my brain understands it.