All Home Automation is local–including the definition of “stable”
One thing that @bamarayne and I have talked about a lot is that “stable” means different things to different people. Or even just at different times in your life.
Before I got sick, I was a network engineer. Crawling under a table and recabling stuff, popping the batteries on a couple of sensors, pouring over the logs and trying a different configuration wasn’t just my job – – it was fun. Troubleshooting itself was satisfying. So my definition of a “stable” system was one that I could mostly keep running even if it required an hour or so of maintenance every week. I never thought about the maintenance time as a cost of the system.
After I was in the wheelchair and lost most of the use of my hands, all of that changed. I can’t do a battery pull by myself. Just getting to the screen with the log entries takes about 10 minutes. And then I have to listen to what’s there. I still enjoy solving a problem, but it better be something I can do with my eyes closed and without requiring the use of my hands.
So now I look at what consumer products designers call MFOP. Maintenance Free Operating Period. Most consumer products aim for a 12 month MFOP. Certainly at least six months. Maybe there’s a firmware update in there every once in a while, but it should be something where the consumer doesn’t really have to do anything except wait for it to finish. Eyes closed, no hands. 
In my 2 1/2 year experience with home automation, quite a few products hit that target. Including the hue bridge, Logitech Harmony, echo, HomeKit.
SmartThings doesn’t come close. Even though these days most of the problems are minor, it’s rare that I go more than a week without running into something that requires more than the “eyes closed, no hands” approach.
So I guess my first question would be how do you define “stable”?
If stable is “90% of my scheduled rules run 90% of the time, but I have to spend about 30 minutes a week making that happen” then that’s probably true for most SmartThings customers at this point.
If stable is “99.99% of my scheduled rules run 99.99% of the time and I don’t have to touch the system for six months at a time” then that’s probably not true for hardly anybody.
So “stable” and “six-month MFOP” are two very different things.
Counting transactions–or days?
Also there’s the question of what “most of the time” means. Technical people tend to want to take the total number of transactions and then look at the total number of misses and come up with a percentage. But that’s not how consumer products are typically rated. Instead, you look at the number of days that are error free, and take that as a percentage over the number of days in the measurement period. That is, a thermostat might do 1000 transactions per day, but if it noticeably fails once every day, the consumer will rightly judge it as broken even though it might have a transactional success rate of 99.9%. (Its MFOP, in contrast, would be 0. 
)
Plan B
Finally, there’s the question of what the Plan B is. Take that thermostat that gets almost all of its transactions right, but still noticeably fails once a day. Maybe all you have to do to fix it is to walk over, turn the dial to the right, then turn it back to the setting you want.
For a lot of people, that would be so minor they wouldn’t even think about it. It would be annoying, and they would notice if a new model that they got later didn’t have the problem, but the fix would just become a habit. You feel cold, you walk to the thermostat, you twitch the dial, done.
for me, I can’t walk, I can’t reach the thermostat, I can’t turn the dial even if I could reach the thermostat. And my service dog can’t do it either. And if I do get cold it increases the frequency and intensity of muscle spasms. So that daily failure would be way more than just a glitch at my house. It would be something I needed to fix right away, and that would need to stay fixed.
So there are multiple factors that go into evaluating what “stable” means at one house versus another. How much time and energy are you willing to put into troubleshooting, and how frustrating do you find that process? One man’s glitch is another man’s catastrophe. 
Knowing what your own level of tolerance for glitches is, or to turn it around what your own minimum MFOP is, is the first step towards being satisfied with whatever system you end up with. 
