Challenging is probably better choice for a word then complexity. Anyway, there are a few different kinds of complexity.
One comes from simply the problem at hand being large / complex and that's something that might be really hard to come up with.
However, it can also be challenging because of unfamiliar tech stack. Be it a new language, new framework, etc. Going too far in this direction can lead to projects being too hard to complete so you want to balance it with familiar technologies as well. For example, if you are familiar with Python and Javascript, picking Elexir and CoffeeScript might be too much. However, Python and CoffeeScript would make a good combination. You get the idea.
I get what you mean - side projects are notorious for never getting delivered. However, there is tons of things that can be learned from that as well. Balancing new and familiar technical choices, creating good tasks and assigning them to milestones, etc.
Noticing frustrating bits and lack of better options is a habit, I believe, which has to be developed. It's easier to notice horribly frustrating things and that's what startups often do, but that might be out of scope for many side projects. However, noticing small things that might be fixed in an evening or a week - that's a habit. I find it similar to the 'I want to blog, don't know what to write' issue.
I get what you mean but I believe the post describes a different issue and talks about side projects which are very different from paid gigs.
In paid gigs, you have hard or soft deadlines, commitments, etc. But sometimes you just want to code with no pressure, no deadlines, no managers, under total freedom and the money is not the first (and often not the second) priority.
In a recent humble bundle, I found Automate the Boring Stuff with Python [0] which partially answers this. After going through the basics of Python, it describes building small projects such as scanning an Excel spreadsheet and sending email to people who are marked as 'not paid'. Projects like this give a different perspective to programming. Programming is not only about building grand projects. It can also automate every day stuff which you wouldn't normally call a programming task.
>>> contribute to an open source project
It's a very very common advice and I believe a wrong one, especially to novice developers. It's simply... Not that easy. Yes, there are tons of projects but most of them are pretty damn large or mature enough so that easy tasks have been a long time ago. Some projects (e.g. servo) have web sites dedicated to new contributors where you can filter issues marked as easy but there are not many projects to choose from. You can look for small, less popular projects... But that's not an easy task either. Plenty of experienced developers have an issue 'want to contribute, don't know where' and they find it tricky, it's much harder for the novices.
One comes from simply the problem at hand being large / complex and that's something that might be really hard to come up with.
However, it can also be challenging because of unfamiliar tech stack. Be it a new language, new framework, etc. Going too far in this direction can lead to projects being too hard to complete so you want to balance it with familiar technologies as well. For example, if you are familiar with Python and Javascript, picking Elexir and CoffeeScript might be too much. However, Python and CoffeeScript would make a good combination. You get the idea.