Dang it, I said I wasn't going to make this all about Browserstack. Oh well. For those curious, yeah, it's a tool I enjoy using. It's not perfect but it definitely made clear how much I liked it when I had to do without it for several months because of a change in focus and perspectives shifted us to different products. Suffice it to say, we are back and I'm happy about that.
When people talk about projects that are small, what they are actually aiming for is "something simple that we can do quickly". However, the problem often arises that scope rises in the allotted time frame. Thus, the trick is to make sure that the systems are in place to be able to keep everything in place. Small projects are risky projects, so it's vital that we trust our team members. We also have to realize that there is no way we will be able to do everything we would like to do. Perhaps swap real-time paring with your developers over having to write bug reports and waiting for a response. Get away from the time killers and project killers; avoid politics, get away from micromanaging and try to not be so beholden to perpetual status updates. Be aware that there is a freedom that can be had with the right attitude. You may not be able to do everything but it's better than doing nothing. Start there :).
Very often, the problem that we have to deal with up-front is the issue of testability. If we are coming into a new QA environment with a lot of manual testing, if the goal is to automate (and even if it isn't) placing an emphasis on testability early can reap large dividends.
Optimizely crowdsources their QA to their entire engineering team so that all of the testing efforts don't just fall on the heads of the testers. More to the point, rather than to have one person testing a hundred commits, having the developers test their own commits speeds up the process and lessens the load on the testers so that they can focus on more pressing areas.
Continuous Integration, Continuous Delivery, and Continuous Deployment are areas that will take time and care to set up but by taking the time to do it it makes it easier down the line. Ultimately the goal is to improve the feedback cycles and build out the infrastructure in a way that allows for repeatability, testability, and implement-ability in many places.