The audience should notice when there is no reverb or delay on the vocals, but not perceptibly notice when these effects are applied. Enhance with effects, don’t over-compensate.
Of course, there are exceptions to every rule. A good example is the gated reverb on the snare drum in a lot of Phil Collins songs.
Another can be found on “Bohemian Rhapsody” by Queen, which has a heavy flange on the vocals. And that same band’s “We Are The Champions” offers a reverse gate application that lends a great sound to the drums.
So particularly if mixing the same band regularly, make it your business to learn the songs inside and out.
This includes going to band rehearsals, where you can calmly evaluate where effects might add something special, and also talk with the band about it as well. (This might seem obvious, but I’ve talked to so many mix engineers over the years that never bother to attend rehearsals.)
The routing of effects is another key piece to the puzzle. If at all possible, they should be sent from an aux send, a pre-fade effects send, or a monitor send (if not running monitors from front of house). The reason is control.
The band I regularly work with plays in a different club every night. This means that every show I readjust my gain structure, fader levels and EQ in light of the specific club parameters and specified volume limits.
By running pre-fader, I’m able to adjust the input to the effects unit and leave it the rest of the night without worrying about clipping when I push up a lead.
I also return all my effects back to a channel on the house console - again for more control. This provides the ability to EQ the effects and control the output easily with the faders, and I can also turn off the effects when the band isn’t playing.
Running it back to a channel also allows me to easily add an effect in the middle of a song, and to more easily and effectively add it to the mix.