Software has the luxury of being edited, automotive engineering also can be edited pre production anyway, the finished products cannot be changed(Compiled Software, a Vehicle).
At some point it's almost always cheaper to rewrite software from scratch (or replace old software with new software) rather than trying to fix old software. Software tend to turn into junk due to bad design choices from the start, due to code slowly into "spaghetti" due to "technical debt" as more and more features are added improperly and without sufficient maintenance, or because the programming language of choice is getting arcane or encourages the programmers to create "spaghetti code". I don't see the Bitcoin Core code being any exception.
Bitcoin is also a protocol. There has almost become a cult movement that changes to the protocol makes older clients incompatible ("hard forks") should never occur. The block size limit has been discussed for many years, perhaps we'll get an upgrade in November, perhaps not. Now, of all protocol improvements requiring a "hard fork", increasing the block size ought to be the least controversial and easiest thing to do - considering how difficult it has been to move this one, I don't have much hopes for doing other changes to the protocol.
Bitcoin has numerous design weaknesses and problems (personally I think the mining centralization is possibly the biggest problem) - SegWit was an attempt on solving some 3-4 design flaws. Well, I'd say it's more of a work-around than a fix. The main selling point of SegWit was that it's backwards compatible ("soft fork"), as such it's complicating the protocol as well as the software, generating "technical debt".
Everything considered, from a technical point of view, Bitcoin will never be able to compete with the best of the alt-coins.