I'll keep additional spies 1 turn away from the key city in question, or in proximity cities for future revolts. What to do when a spy fails is really situational. Often with two movers I'll just go ahead and move onto the next city I have a spy building up the wait bonus for, and replace the failed spy in the meantime. But depending on espionage points available, defenses, the importance of the city, and the importance of speed I've done everything from attack the city with big losses, "cold" city revolt, wait 1 additional turn then city revolt with new spy, wait for full 5 turn wait bonus, or destroy walls instead.
Stacking spies will greatly increase the odds that they're discovered.
As I understand the odds of discovery increase PER additional spy (@Kaitzilla
can probably confirm). This makes it very difficult to accrue wait bonuses consistently. It's easy to fall into the trap of using more spies = get more caught = use even more spies next time = have even more get caught. If it's the end of the game tank war, sure, throw in a stack of 10 spies. Earlier than that I'm dealing with 1 spy per city, but I put a lot of effort into advancing them ahead of time and having backups ready to swoop in.
Another minor point is if you're able to attack a city turn 1 of the war, attempt the city revolt BEFORE declaring. It'll have less chance to be caught and if it is caught, can just delay the war a minute.
I will usually aim for building up a decent espionage surplus on the target ahead of time, essentially banking in 1-2 "fails" that result in me having to rush some revolts without the wait reduction. The benefit of this is that the extra espionage can give you full vision of the AI. This often allows me to pull away my emergency defenders instead of aggression, because I can see exactly what the AI can and can't do.
There is definitely a possibility of really bad spy luck derailing an attack that's dependent on them, that is the tradeoff. I would say with a traditional HA rush, I often just build 1 spy to utilize the free EPs on a city revolt or a few wall destroys. Can be hugely worth it, and if it's discovered you can just attack like usual. There are a lot of UUs that people think are underwhelming because siege does the heavy lifting, but if you use spies with them they are actually very good (chokonus, samurai and cataphracts come to mind). I've even had some success with praets, berserkers, gallics>maces and all the gunpowder UUs. The potential advantage you gain is from having to spend fewer hammers on spies than siege & potentially be faster, at the cost of your economy spending. This eco spending is mitigated somewhat by having some free EPs accrue, not needing to tech construction/engineering, spies being the 'best' gp to run when relevant, and being able to pay for that economy cost during the war while your troops are on the way.