My solution to E: Double Permutation Inc. was similar to the reference solution (except I used sets rather than maps), but it turns out that calling .size on TreeSet.tailSet() is slow ($$$O(n \log n)$$$) and caused TLE. I submitted that in the last minute and would've got the T-shirt if it passed. Oh well...

After contest time, I submitted a solution using BIT instead and it passed easily, but good to know there is a solution that doesn't require non-STL data structures.

For F: kotlinkotlinkotlinkotlin..., isn't Fleury's algorithm too slow ($$$O(n^2)$$$)? I spent too much contest time implementing Hierholzer's algorithm from scratch as I never used it before. Handling linked-list-nodes is just so error-prone. However, after the contest time, I discovered that a mere ArrayDeque can also implement Hierholzer's by "rotating" the deque whenever you get stuck. (The deque then may need to be rotated again after the path is complete, so that it starts with vertex 0. It can be shown that you'd never need more than $$$n$$$ rotations in total.) Example code: #60237987. Great learning experience though.

I think the should be a small correction in editorial of D. It should be a>=b and not the other way round