I have been playing Shadowrun Returns (on linux!) recently, and enjoying it for the most part. There are a few things that stick in my craw, though.
Nothing says "failed op in progress" like traipsing into a highly-secure corporate office without so much as a floorplan. I like heisty games, where foreknowledge is key. Most missions in Shadowrun Returns basically require that I go in blind and just shoot the place up with so little concern for opsec that my team is bare-faced and has mohawks up. I presume that this is to make identification from the security footage easier, and thereby generate more gunfights, which are Fun.
This is a less than satisfying state of affairs.
Also: you got your fantasy in my cyberpunk. Do not want. Likewise, their hacking is adorable. I recognize that it was added as a last-minute feature, and it's a perfect gameplay fit for a turn-based dungeoncrawler, but oooh man it's also terrible.
All this has me in the mood to roll myself a proper cyberpunk rpg. The basic premise of Shadowrun, that the megacorps need dirty deeds done by deniable assets, is great. I'm disappointed with the implementation, though.
What's more, heist games tend to not go over so well with my group, which has 1) a fairly limited patience for planning, combined with 2) a collective indecisiveness which tends to generate muddled and incoherent plans even after some time planning. Halfway through the operation someone will go "wait I thought you were supposed to have done X ten minutes ago." I have been that someone; I have also been that you. So running a heist straight, in a perfectly chronological fashion, doesn't work so well for us.
Might be time to take a note from the movies, where things often cut back and forth between the planning and the execution. Instead of a planning phase, followed by an execution phase, we might have an intelligence-gathering phase, followed by a combined planning/execution phase. I see this working like so:
In the intelligence phase, the PCs take actions to gather some numerical manner of 'Intel Point'. Actions which earn intel points include getting floorplans from the building inspector's bureau, hackin' around the intertubes, talking to hardware suppliers, maintenance contractors, and former employees of the target, dumpster diving, getting an employee drunk and pumping him for information (and a covert scan of the RFID tag in his ID badge), seducing the plant manager's wife, and all that fun stuff. The more groundwork you do, the more Intel Points you have to spend later when things start to go wrong. On the flip side, if the target catches wind of your intelligence-gathering efforts, they're liable to tighten security preemptively; put up an extra checkpoint, add metal detectors, double up the watch schedule, change all the passwords, keep a rapid response team in a black van on call, maybe shell out for that cavity-searchbot the chief of security's had his eye on for a while now. The usual. If gathering intel goes poorly, the number of surprises you wish you had intel points to preempt will rise. If intel gathering goes particularly poorly, maybe one of your guys ends up ID'd or caught, and then you're really in trouble...
At some point you have to make the call and go for it; either your intel gets stale and starts to go bad, or you have a deadline from your employer, or the risks of continued intelligence gathering just aren't worth it. So you formulate a preliminary plan - something like "The face is going in for a job interview under an assumed name. He will 'get lost' looking for the bathroom, and then open a back door for the muscle and the ninja, who 'borrowed' a plumber's van and some overalls. The ninja will proceed to the objective while the muscle and face provide distractions or covering fire, and the hacker provides security camera and comm interference from some local university's library wifi and a spoofed MAC address."
And then things start to go wrong, and the players start to spend intel points to alter the plan around these new obstacles if they can't think of a way of dealing with them on the spot. The rule, though, is that when the obstacle is revealed, you have to choose immediately whether or not to revise the plan before you take any actions to circumvent the obstacle. If you planned for it, you never met it; permitting actions gets you into an even weirder hypothetical state and basically gives you two shots at each problem. Intel point costs are based on the relative secrecy of the obstacle; it's not a huge leap of logic and planning to assume that a secretive installation might have guard dogs and to plan accordingly, but it is probably harder to find out beforehand that the thing in the top-secret lab is actually an angry tankbot, and then even more difficult to smuggle antitank weapons into the facility.
I dunno. It'd be an interesting experiment with a flavor of party-shared associative action point, and perhaps a solution to our heist-game woes.
If I were to really build such a game, I think I'd want to do it on a Traveller chassis, maybe with a slightly wider die spread (2d8 or 2d10, with 10+ and 12+ for success respectively) so that modifiers could be a little more granular and so a +4 modifier isn't as game-breakingly good. But Mongoose Trav is about the right grade of complexity, and a lifepath-type character generation system would be perfect for cyberpunk. Everyone's down and out, but not necessarily for the same reasons. Likewise, Traveller has good support for dangerous performance-enhancing drugs, the law level mechanics would certainly be hands, their cybernetics book was reasonable, and there is no prior expectation of elves and mages. Hacking would need some work, but it always does.