I get 141.5 seconds for the first time, (laptop in suspend mode during the second), 84.8 seconds for the third, 67.9 seconds for the fourth, and

  re2/re2.cc:821: DFA out of memory: pattern length 2493008, program size 1323834,   list count 1089251, bytemap range 54
  is concatenation? True time: 102.0 seconds

for the fifth. It would appear to not be caching, or at least not in a way that makes this timing faster than the NFA.

In any case, my point is that we know this problem can be solved in O(n) time.