There are many test cases related to nanosleep/usleep function. They test the performance of timing roughly similar to the following code:

gettimeofday(&start);
evutil_usleep_(100000usec);   // sleep 100000 microsecond (or 100 millisecond).
gettimeofday(&end);
diff = end - start;
assert(80000 < diff && diff < 120000); 

This code runs successfully on my own computer (Linux, Windows and Mac).
On travis-CI, it only succeeded on Linux and Windows, but not on Mac(The diff is likely to be greater than 120000).
In travis result, there are 133 failures, 124 of which are sleep related.
I guess the issue is caused by the Mac virtual machine which travis used. But there is no evidence to prove it.