After some TLEs, I submited a source which just reads data and I found that there are over 8000000 datasets.

The time limit is set to 3 seconds, not to 8 seconds as reads in the problem statement.
8000000 datasets were read in near 3 seconds, but there are more datasets.

So, this problem is "locked" by the problem setters?