200: simple seconds to h:m:s. 2 consecutive mod 60 and div 60. No big deal.
550: simple DP-like process. A for loop: deducting one bit each time. Point is corner cases.
1100: special backtrace on a tree. Direct thought like Greedy backtrace is not optimal - because it only needs a final optimal cost - no trace need. So.. sum_all_duct(2*duct_cost) - cost(deepest-path). That's it. Because the deepest node is the last to visit, and then we dont' need get back to root. WOW.