Here are some things I learned while trying to run a heist in my old school game:
Reconnaissance is Exploration
And much the way you would expect it to take some time for players to get the feel for a dungeon level, it will take some time for them to case the location of the heist.
One of the reasons it takes so long is that players will adjust and concoct new plans for each bit of intel they get. Find out about the guard patrols-- plan all the ways they might be distracted; realize there is a sewer underneath-- plan how to use it stealthily. So, learning about a place will be a continuous, rolling strategy session.
That's all fine, because with the right players it's great fun, but it takes time. Next time I would plan for the fact that casing the joint will require a whole session and the heist will go down the next session.
You could give them all the intel at the start, but then I would treat that the same way you would treat giving a party a map and rundown of a dungeon level: make the most of the player's anticipation of very difficult parts and make the most of the confusion and drama a few errors thrown in can create. So the fun shifts from, "What could be behind this door?" to "How are we going to handle the troll behind this door?" Or, even "Why did the thief lie about the troll behind that door?"
Consequences, Consequences, Consequences
The whole reason you case a place is because it's too dangerous to go right in. That may be because you will get killed or the person you are trying to save will get killed, or maybe, the object you're trying to get to will just teleport away.
This seems completely obvious, but it's something I messed up. First, the only legal repercussion my players had faced was not that bad. Second, I didn't put enough guards at my location. Third, I didn't limit my players movement/power enough.
So I recommend: Make the heist happen somewhere where weapons are forbidden. Make the heist happen where trouble is expected and if the guards come the players will likely be killed. Or make the legal consequences dire. And make sure the players know all this. In other words, if they can just enter the place, swords drawn, shields raised and win the day, it's a dungeon, not a heist.
I put a lot of effort into not trying to expect what the players would do. I put trees so they might climb the walls. I put a sewer they might sneak through. I had pilgrims flowing in and out of the place so they could do some easing reconnaissance and possibly sneak in that way. I put a place they might dig under the wall. I included a supply/corpse wagon that they might sneak onto. I also made sure the party had several magic items that might be of use.
And this worked pretty well. But, I think to encourage players to break into teams, especially, more than two teams, would require more skillful design.
Why does it matter? Because much of the drama of the heist is relying on the actions of your compatriots. If they screw up, your job is that much harder. Also, it puts players in the position of an observer for a bit, rooting for the other players in the way a regular dungeon adventure doesn't.
So, I would want to design the heist location better to try and make a two or three prong approach a necessity. Magical locks, magical alarms, nested locations.
The Trouble Meter
I got two results of complications during play. They both worked well just the way a random encounter mixes things up in an interesting way. But, the whole Trouble Meter went out the window because my players immediately made themselves obvious to everyone by creating a loud distraction. I never ended up escalating the trouble meter.
So, you need to take distractions into consideration, because they are part of the genre. I think this might be related to me not making the location hard/scary enough. It's one thing to have a patrol distracted for a bit so someone can slip by, and another to say "Hey, here we are. We are going to cause a hell of a lot of trouble!" Isn't that why guards are there? My players should have been much more leery of raising the general alarm. Or, I should have had a segmented location, where the second team could have still raised an alarm within their area.
The Event Clock
In a pseudo-medieval setting you have dawn, dusk, and noon. Anything else will have to be rung from a church bell or something. Times won't be as precise as in a more modern setting. So the clock wasn't that important during our session.
I think this is closely related to teams, though. One of the basic ways one team can let down another in this genre is if they get their timing off. They might get hung up because of a complication or get sloppy and start a distraction too soon. So, to make the clock more relevant you need teams and you need time dependent tasks: "Once the magic door is opened you have thirty minutes before it closes again, permanently."
But, it means you need a way to measure time too. pseudo-medieval characters aren't running around with Timexes. How do they know when they are supposed to start the distraction? This means players will need to utilize magic that extends the senses ("When you hear us start yelling, go!") or creates a link between the players. Or, that the location itself needs to offer some semi-regular events that players can utilize as time markers.