Megalomaniak Unless there's something you REALLY need from a new version, minor versions included, it's typically not worth the risk to upgrade if what you have is already working, especially for large projects. Upgrading for the sake of upgrading has a potentially large technical cost for minimal benefit...if you started with a particular version, you presumably did your due diligence and determined that it was suitable for your project, so there usually wouldn't be a compelling reason to upgrade anyway. The exception is patch upgrades which contain no new or altered features, just bugfixes.
If you do decide to upgrade the engine for whatever reason, you would do so with the expectation that some things may break and you will need to allocate development time for it (and of course you'd do that with a separate branch of your project in case you need to revert). If there's some sudden new requirement that requires an engine upgrade to continue publishing new updates of your game (e.g. Apple changing stuff), it would usually be better to fix that one specific thing in the engine source code yourself instead of doing an entire upgrade.
Remember, I'm pushing back on the idea that "maybe you should only make small short-term projects with Godot because updates break stuff". Long-term projects are fine; there's no mandate to update the engine mid-project and in the industry it's generally discouraged. Upgrade to the latest when starting a new project, sure.
