In my work as a scientist, I work on projects of a range of scales. While my major projects, which I do for work, are generally multi-year and often team-based, I also frequently start smaller projects. These are generally done to scratch an intellectual itch: quantifying a certain effect that I've been pondering, wondering how a problem looks when put in a different framework, etc. They are rarely team-based although I may have collaborators that provide data and hopefully some feedback.
My smaller projects are almost always computational, as they're cheap and low-overhead compared to field-based projects. Field mapping is also cheap and fun, assuming the site is close to home, but requires a big block of time; additionally, in the US, geochronology is almost always necessary to get new results and that's expensive. There is a lot of fertile ground in computational geoscience, particularly with discrete mathematical approaches as opposed to continuum mechanical approaches; the former are also much easier to code and the programs can often run in seconds on a laptop, which is not the case for finite element models of orogenic or earthquake dynamics.
Projects that take longer than an evening fall into either the side project or sabbatical project scale. Side projects can be done in evenings and weekends over a week to a few months. At this point, after about a decade of modeling, I almost always have an idea of my approach before I write a line of code, and I can usually get a prototype up and running in a few hours or a few nights. This is enough to validate the idea, and if it holds up (it often doesn't), I know where to go. Then if it's good, in a few days or weeks I'll be able to get to a point where I can start writing it up for publication, if it's that sort of project. Alternately, I can write it up as a blog post, or just incorporate the results into my mental model of the world or add a new tool to my modeling quiver or whatever. I find working on side projects, this intellectual exploration and itch-scratching, to be quite gratifying and I have learned as much or more from them as from my formal education and paid work.
Writing and revising is the most challenging part, because once I have the results, my interest really starts to wane and I am ready to move on. But in order to get any return on my time investment other than personal intellectual development, I have to publish, as much of a drag as it is. I have found that I get no satisfaction from publishing, and peer review is on net a negative–reviewers rarely invest the time to understand a paper that is not exactly what they do, and it's clear in the comments when a reviewer read through a paper just once, which does more harm than good as respectfully rebutting poorly-reasoned comments takes tremendous mental and emotional energy for me. However one out of every four or five reviews is actually helpful and improves the work. Nonetheless, no one will take my work seriously if it's not peer-reviewed (quite reasonably), so it must be done for certain projects.
I generally have 3-4 side projects going at once. They are often peripherally related to my main job but not part of my paid work. They could be completely different, however. As some examples, I've been working on stress estimation and statistical paleoseismology through strings of side projects for a few years. The stress work started as a side project during the end of my PhD, became my post-doc project, and has continued with marginal progress for the subsequent 4 years. The statistical paleoseismology work has yet to be associated with paid work but I suspect it will become central to my work in the next few years. Both of these are at this point strings of smaller side projects that are related but can stand independently, rather than longer-term indivisible projects. I have also done projects such as the slip rate calculator or the tectonic plate motion viewer that are one-offs.
I tend to cycle through side projects as interest or inspiration dictates, unless I'm in the writing/revising phase where I really have to focus otherwise it won't get done. I might work on one project on Tuesday and Wednesday night, and another project on Saturday, or whatever. I'm a restless person and need to do this to keep from jumping in my truck and driving dirt mountain roads for the rest of my life, but I don't think I know anyone else who does side projects at this rate, or even close.
Sabbatical projects are those that take basically full-time effort for 3-4 months to a year (i.e. 300-2000 hours of work) and are enough of a departure from existing work that a lot of new foundations have to be laid before any meaningful results can be derived. I have started several of these but they always get abandoned. I have too many ideas to just focus on one project in all my evenings and weekends for a year or three. I don't always know that a project is this scale when I start it. Sometimes, after starting the modeling for a side project, it will not fail but not succeed, and I understand it will take a lot of work to get anywhere. Estimating topographic and tectonic stress fields in orogens worldwide is one such project, as is developing a graph theoretical treatment of geologic data (which I think is a natural and powerful way of encapsulating or encoding data with geologic relationships, as well as computating geologic histories, stratigraphic correlations, etc.), and tectonic geodesy that incorporates faults and is in 'Euler space' defined by the relative rotation between stable blocks. And many others, I'm not lacking in ideas (not to say they're good ideas). Other bigger ideas in the post-it note stage, such as a thorough treatment of earthquake clustering that combines new statistical measures and physical modeling, or a much improved understanding of how faults are loaded (do faults really have big creeping sections at depth that load the upper, locked part, or does the loading happen in the upper crust and the ductile down-dip extend is primarily deforming through post-seismic processes?) can hopefully be subdivided into side projects or made quicker through collaboration. I have to be very careful not to start putting time into these projects, once I recognize their scale.
The hard part about sabbatical projects is finding the time, i.e. the money, to work on them. I'm not a professor so I don't get sabbaticals, and I'm terrible at getting grant funding. Both me (a semi-independent, non-academic professional scientist) and my ideas (roughly-sketched, often well off the beaten path, and with a high risk of failure) are pretty much the antithesis of NSF's ideal, as NSF basically wants to fund graduate students to do low-risk science. This seems to suit both the universities and the government but makes it tough for me to get like $50k to try something new. But given the nature of these projects, nothing will get accomplished without a solid investment.