Software needs to maintained, patched, backed up, verified etc. It has bugs, security issues, hardware breaks in weird ways. This takes time and skill - ideally you'd need two or three people that are capable of fixing problems with the install. (one ill, one on vacation, one available). This is something that detracts from the actual work you're doing. I'm very much an ops person and I actually like tinkering with a gitlab install - it's just so many moving parts that I prefer not to run this for my company since it would eat a substantial chunk of my time just caring for this.
I note that the Linux kernel lived with bare Git for many years.
The Linux kernel is a very specific case with a very specific development model that likely doesn’t apply to most other projects.
You need to compare the cost of self-hosting to the cost of SaaS - INCLUDING the risk of getting locked out.
One downside of the SaaS model is that you are just a very small customer in the bigger scheme and they can't really justify spending money on servicing you. Let's say you are company of 5 people, paying 50 bucks a month for a service - how many hours per year can they spend on servicing you before you become a net-negative account? You much power do you have in a negotiation if you are a net-negative account?
It probably isn't sustainable for a business to only consider this aspect. One thing that comes to mind with companies that thrive with a large number of small non-B2B customers, who individually don't tend to have much power, is that they understand that people love to talk about customer service when it's bad, and occasionally when it's very good as well. Word spreads, and nearly everyone places at least a little weight on this public perception of kindness or flexibility with customers especially when it isn't in the immediate financial interest of the company to do so.
As far as maintaining the system is concerned: setups that are hosted by 3rd-parties also need maintenance. Someone has to understand how it all fits together, and how to fix it when it goes wrong. So you still need a team-member working part-time on SCC, CI and deployment.
If no one in your company can do that.. hire or outsource.
(I have been doing it for years)
For example, the source code as well as the tickets around a software tend to be the most critical assets of a company. As such, you need one or better 2 systems to host the source host and ticketing. However, such a system needs backups, so suddenly you need to maintain a backup solution, you need to implement and monitor the backups being created, you need restore tests. You end up needing some kind of monitoring as well. As well as 2-3 dudes at least part-time maintaining all of this capable of replacing each other during sickness and vacation.
That's a lot of stuff as well as a lot of manpower as your base cost. Of course, once you have that, you can self-host a lot of things easily and maintain excellent uptime at minimal risk, because these base services scale very well in complexity. For us it makes sense to do this, because unplanned outages at 100+ developers are seriously expensive and risky.
However, if you have 3 developers and a clock ticking to find product market fit, you don't have that budget - or spending it this way does not make sense. So you buy.
