The ability to ssh into an existing proccess is great, and let's you do some of the things that
heroku run
lets you do. (And some things it doesn't, becuase heroku run doesn't actually let you connect to a current process).
But the main difference is that
heroku run
will start a
"dyno" for your
task. Not use an existing web or worker process. It boots it up on demand, and charges you for the minutes of use.
This is useful becuase you can run things (or open an interactive console) in a separate environment from your actually running production web/worker, without worrying about interfering with their RAM or CPU resources -- to run a temporary one-off task, or an interactive console.
In terms of render's model, I could imagine this working by defining a new service type with it's own "build command"... but then I guess you'd still need some CLI way to launch one of those with a custom one-off "start command".