we've been experiencing ghost-in-the-shell kind of behaviour last couple weeks where we'll save a project at the end of the day, come in the next day, and have random parts of the project 'mutated' back to either previous versions of the project or sometimes just random stuff will change: flags flipped from false to true (particularly visibility), or tab name rearranged.
the unsettling part, is most of these changes are things we haven't touched in weeks or months. the other thing that causes me great concern is the three versions of each team member working on the team shown as having the project open... when only one or none of us has the project open, as noted below.
if anyone is having this kind of issue, i'd love to hear if you've managed to find the reason for it and whether it's fixable. it's causing a LOT of problems, as you can imagine.
A gateway restart should get rid of those pesky ghost users. As far as I've been able to discern, they apear when a designer connection to the gateway is improperly severed. I guess the gateway has no inherent way of force checking if sessions are actually still active.
This specifically happened to me when I closed my laptop to go to lunch, when I came back there were suddenly two of me.
As for why your projects seem to be "mutating", I wouldn't say its impossible that these ghost users have some sort of content lock, on certain parts of the project, that seem to override yours, though I have no proof of this.
yeah. problem is, we can't be restarting the server every couple hours. which leads me to what i've always suspected the core problem is: garbage network connectivity. but beyond that, the presence of the ghost users causes me concern. as you said, i suspected they existed for the purposes of caching content until reconnected... but they aren't disappearing when reconnected. in fact, i don't even notice that there was a disconnect to begin with.
i'll dig into the network side of things and file a ticket with tech support. maybe they can shine more light on this because, as it stands, this is very problematic. we're basically restoring end-of-day backups the next morning, every day. not fun.
I vaguely recall reports that some brands of routers (or versions thereof) would kill off websockets that survived more than 30 minutes (or something like that). That is, it might not be conventional network breakage, but bad behavior along the path.
i would not be surprised if that's the issue here. we have a literal dinosaur of a Cisco that desperately needs replacing. this might be the ammo i need to push that over the edge (literally, if they'll let me). thanks for the heads-up on that. that just saved me hours of generalized combing through my running Wireshark capture. you da boss, Phil. off i go to dig through hardware specs and do come specific testing to confirm if that's the case here.
EDIT: blow me down, that's the answer right there. our Cisco shuts down web sockets after 30mins of inactivity. unless you turn the firewall off. gotta love it. time for a new Mikrotik (yaassss....).
off i go to dig through hardware specs and do come specific testing to confirm if that's the case here.
unless you turn the firewall off. gotta love it. 
time for a new Mikrotik (yaassss....).