Didn't know about this. What's the difference between a new dyno and an existing one? Why would they act differently?

They don't act differently, its just a dyno separate from the web and other processes.

That's good for things that change shared state, but you can't do anything that would alter a running dyno or interrogate its state, unless you built your own hooks or something. But building something like that would kind of obviate the perks of a NoOps system.

The changes you made by running 'heroku run bash' is temporary and can be reverted any time.

`heroku run bash` is still the best option, but I've always wanted to ssh into heroku dynos myself, so I did this:

https://github.com/fabiokung/heroku-ssh