Game development meta should be:
1. Decide on the type of games you want to develop (RTS, RPG, advanture, FPS, etc.).
2. Make a game engine for it.
3. Make assets for it.
4. Finall, make a game by combining the engine with assets.
5. If successful, create a sequel/similar game.
This is how all major game-dev companies operate. Indies can do this, too. Just on a smaller scale.
Patentedly untrue. MANY major game companies use existing engines. For example, the pokemon gen 4 remakes are on it.
Depending on what game it is I can see someone using something like Raylib to build their engine instead. The abstraction level there is about the same as Godot. You get high-level API:s for a large number of things and support for many platforms (far more than what Godot or Unity supports). What you do not get is a GUI for building scenes or editing scripts.
Starting from a lower level with something like SDL sounds like something more suitable for hobby projects, like coding a game in TIC-80 or writing a 16-bit DOS game in FreePascal or other hobby-activities (things that I enjoy). Fun, but probably almost never a good idea for a business.
- learning is also fun
- Most engines (by force or by available tutorials and docs) railroad the end user into some specific genres (in my experience: side scroller, FPS). If you wander too far away and understand the risk/complexity, it might be worth considering.
terrible idea.