No minding at all,
@NuttsnBolts. Some could've been avoided, others were unexpected.
Being based on a prisoner / captor concept, the chief design flaw was lack of hope. The architect of the prison had designed it after reading and using the Evil Overlord list. Thus the security systems were well beyond what you'd find in even maximum-security prisons on Earth. In addition, even if they had escaped, we had no system in place for what characters could do then.
Other mistakes we made involved imprecise mechanics. Being a high-end large-group RP, we did not have mechanics in place to keep track of IC time. Not in any good or consistent fashion. The current iteration has 28 characters spread out across 17 different players, so managing IC time is mandatory.
One tip we have is this: never underestimate the value of a good pair of Co-GMs. The larger the RP, the more necessary. People will have problems in real life and that includes GMs. Having a couple of reliable Co-GMs means that at least one GM is available most days.
We've also found it valuable to be critical in the CS process. Not rushing things there eliminates the majority of problems later, be they godmodding/Overpowered characters, players out of their league or earwax incontinence. (ok, maybe not the last one)
Another issue we had was overcomplication. If something in the RP is so complex the GM(s) alone understand(s) it, then it probably doesn't belong there. A timestamp system needs to be simple and easy to understand. If the RP is based after a certain event, it might seem logical at first to use minutes/hours after event, but that stops making as much sense once you get into events on the following day. Most don't want to bother trying to figure out how many days after the event 432 hours and 19 minutes is.
If a captive player population has no hope at all of getting away, it will not encourage them to try escaping.
We think we deviated a bit above, but some of the mistakes weren't easily quantified into a single sentence.