Code must be copyrighted manually as there needs to be official proof of it's uniqueness.
A copyrighted work doesn't have to be unique. It's just highly unlikely that anything creative and substantial wouldn't be substantially unique.
It's illegal to use someone else's copyrighted code, so there needs to be a copy of the code submitted manually in order to ensure that no such thing as a duplicate code can exist without the approval of the dev.
That's a non-sequitur. Copyrighted works don't have to be registered to avoid duplicate works. It is the responsibility of a copyright holder to vigilantly look for possible infringement upon their copyright, which includes things like derivative works. Since registration on works are not mandatory, no registration database could be used to verify uniqueness and since derivative works are still covered under the original author's copyright, uniqueness would not be proof of non-infringement.
Code does NOT count as "works" and doesn't fall under the Berne Convention because it would be jokingly simple to rip the code and use it yourself without anyone having any way to prove your is a rip-off.
Actually, there's been plenty of cases where people have tried to take code and pass if off as their own for which the rights owner has successfully prosecuted their case precisely because (1) most people who rip code are idiots and leave lots of obvious strings to compare against and (2) code->machine code is generally a very match-able thing unless significant effort with obfuscation has occurred. Regardless, the validity of a copyright does not hinge on the ease of infringement nor the difficulty of proving infringement.
Even if you release it online, it's possible to change the displayed date of the host you are using for the code, thus making it impossible to properly enforce copyrights that way.
As a general point, the basis for proving that you're the original owner can be based upon a lot of things. Usually a simple display date is not the cornerstone proof. Having said that, yes, registration is often done on works precisely because it makes it much easier to use a registration database's date as a means of validation. That doesn't mean one can't provide proof otherwise, as generally speaking people don't dump 100% of their development work online which means the incomplete or broken sub-stages can be presented in court as a means of validation. That's pretty compelling proof vs the other side being only able to present later versions of the code as their own or having a fully formed code base as all they've got.
Seriously, though, what you're saying applies just as well to anyone else trying to rip off anyone else, especially prior to the internet and when about the only medium available was ink and paper. If even under those conditions registration was not mandatory, the notion that code is somehow special because of how unprovable ownership could be would apply just as well to, you know, everything else put on the internet or otherwise digital or whatever.
Having said all that, I can imagine that your or other countries have chosen to interpret the Berne Convention in some fashion to try to treat code differently, but that's clearly in violation of the agreement. Which is to say, it's little surprise when countries ignore international treaties they're party of or choose to selective interpret its language. Having said that, again, most countries do abide by the treatment of code as automatically copyrighted.
