Why is the Win32 boilerplate the way it is?

[Jeremy Sorensen via Digitalmars-d-learn]

I found an example of boilerplate code for Win32 programming in D 
here:
http://wiki.dlang.org/D_for_Win32

I have some questions.
1. It appears that the call to myWinMain from WinMain is to 
ensure that any exception or error is caught. At first glance it 
looks like this is to ensure that runtime.terminate() gets 
called, but in fact it doesn't, the catch block doesn't do it and 
there is no scope(exit).  Is this a problem? (And what would 
happen if you didn't catch the exception?)
2. Why does the boilerplate return 0 on success and failure? (If 
the return code is irrelevant, why the comment that says "failed" 
next to the return code?)
3. I can't imagine a technical reason why the myWinMain signature 
has to match the WinMain signature. Wouldn't it be better to omit 
the hPrevInstance since it isn't used? (Or are we preserving 
backwards compatibility with Win16?).

If there is a resource somewhere that explains all this I would 
happy to consult it but I couldn't find anything.

Thanks.