Right now, I've got several local machines hooked up to gitlab.com as runners. A mixture of virtual machines for various platforms, and docker hosts. No kubernetes yet; but likely at some point.
The main problems are that sometimes gitlab.com is flaky. Pipeline jobs fail, never get started, or never complete. Not often, but enough that I can't guarantee things will work without manual prodding. I saw quite a few instances of this a few weeks ago, but it's been OK this week. Stuff like the runner timing out pulling a docker image, the job completing but not actually finishing, or the pages job running but the deploy step getting stuck with no way to debug it.
Other things are UI annoyances, like the pipeline status not updating frequently enough, leading to repeated manual page refreshes, particularly on navigating back in the history to the pipeline page from a specific job page.
I'm really glad to hear you're excited about the way we integrate everything together.. that's something we're really proud of with our product and an area we're going to continue to invest in as well. This year we're going to explicitly plan things out so that we're building a breadcrumb trail back for all users to start taking advantage of those more advanced features.
You can check out the rest of our vision for 2019 here: https://about.gitlab.com/direction/verify - would love to hear your thoughts and feedback on it.