I haven’t been spoiled until now … but I didn’t find the solution either 
cblavier
2
ferd
Author of Property-Based Testing with PropEr, LYSE, & Erlang in Anger
Part 2 was clearly one of the most intimidating problems to me this year. They dressed it up in fancy clothes with a big name and I imagined it to be worst than it was.
Source at day12.erl · GitHub
2
qhwa
After some failed attempts, I realized that it requires seeing the problem in a different view point. Brute force would not work anymore. I didn’t find the solution immediately. It was frustrating but I let it go.
When I had time this evening, I read the data carefully once again, and thought, maybe I can try separating their positions and velocities of 3D into three groups of 1D data, that would significantly lower the pressure.
So I took it a try and fortunately it worked!
Here’s the solution I end up with. I’m so glad I learned this new pattern. 
BTW. It finished part 2 in 580ms on my laptop. Not bad.
3
mortalisk
I had to get a hint about looking at each component of position/vector independently. Then it was relatively easy.
1
cblavier
I did it !!!
I’m proud to have figured out the one axis + LCM trick by myself (even if it involved not sleeping anymore after 5AM, damn brain … 
)
I had initially part2 solved in 6 seconds. Running each axis in parallel tasks allowed me to get result in 3 seconds. Then after reading @qhwa code (good job btw 
), I simplified my data structure and switched from a map of structs to an array of tuples.
Part2 is now running in 800ms 
, my code: advent/lib/2019/day12 at master · cblavier/advent · GitHub
I’m one day late now … 
2