Here I usually program the first thing that comes to my mind. Most often this is the simplest, most inelegant logic - in my defense, the tasks get published at 6 AM in my timezone.
Typically, I don't go back to fix issues unless it's incredibly ugly and it really bothers me.
Also, in most cases the solution to part 2 overwrites (and destroys) the solution to part 1. The commit history should tell the story for part 1, although I might go back and organize it properly after the event ends.