For example CDPR has their own engine RedEngine which basically allows you for completely seamless open world without any loadings and at the same time allows for very detailed indoor spaces, also without loadings.
I've watched a presentation by CDPR dev from some conference on how they optimised their open world.
The engine itself doesn't contain any advanced optimisation algorithms to handle open worlds and is extremely rudimentary in that regard.
They had an external tool that run on a server and constantly analysed the current build worldspace with refresh time of few hours. It was generating statistics of memory usage, lighting, polycount budget, etc. for every spot in the game.
Reports from this data could be displayed in a graphic form of a gameworld map used to find badly optimised spots.
Whenever devs testing current build encountered bad performance spot, they could use informations generated by the tool to figure out what is the cause and manually optimise assets, lods, lights and write streaming engine scripts for surrounding area.
For the next game they planned to upgrade the tool to automatically perform the most mundane worldspace optimisations.
You could drop the same army of underpaid eastern european junior developpers, to work alongside level artists and manually optimize the shit out of every single spot of the worldspace, on current UE4, or Frostbite and achieve even better results most likely.
The previous iteration of Red Engine used in W2, had tiny levels separated by pairs of doors and corridors, or rocky paths with a pair of climbing spots, to hide loadings ME elevator style.
While the character couldn't jump and was glued to the walkmesh and they were so proud they can have a bridge you can walk under ( but not jump from, unless it's a scripted animation) unlike with Aurora Engine.
They were so proud of these magnificient developments, that they even used them as marketing bullet points
Think where their games codebase could be now, if they could start with current UE instead of developing such barebone patchwork of an engine from scratch.
When they started UE wasn't what it is now, neither licensing was so cheap. But these days priopietary engine makes sense only if you are big enough to fund it with less than 5% of the revenue, or the game idea has a very basic requirements in terms of engine features.