Cactus Music
Still monitor direct at the interface. Now turn the effect send on the track up full but turn the tracks volume way down until you don't hear so much delay. Kind of a hack but it does work better than hearing the delay.

Putting the reverb (or a delay it that's what's needed) on a pre-fader send can simplify things even more.

Enable input echo on the vocal track then turn the track fader down and adjust the send until the singer gets the amount of reverb they want. Then afterwards turn up the fader, switch the send to post-fader and adjust as required to balance the recorded track.

It's also a good idea to temporarily Archive all the tracks that you don't want to hear while recording. All audio and midi data associated with these tracks will then be ignored by the audio engine.

I just use the zero-latency monitoring feature of my MOTU.
 

I use the monitor system in my RME audio card to mix the live audio back into the headphones..no latency.
 
Before that I would use a mixer that would send one track into the audio card and split another into a mix.  If you mix the vocal track back into the mixer at a lower level you can put reverb on it and the vocalist will hear the vocal with a little delay and verb on it.

My "Mixer" IS my audio interface so for me this is not an issue!  I can do full reverb, or any other effects (built into the "mixer") without having that come into the recording...