Fix coverage in loops

[nohwnd]

Indeed, the new Profiler is much faster than the breakpoint method. Our unit tests now run in about 10 minutes versus 17 minutes. However, I am seeing a few differences in coverage between the two methods. After processing the coverage file with Report Generator, I observe the following discrepency (on the left is the coverage using the Profiler while on the right is the coverage using Breakpoints):

What the profiler is reporting for these lines doesn't make much sense.

Originally posted by @fred-gagnon in #2052 (comment)