Your solution has a bug and works incorrectly for some inputs, depending on order of intervals. See these intervals which should sum up to 19:

[(1, 6), (1, 5), (10, 20), (16, 19), (5, 11)] - your result is 20
[(1, 5), (1, 6), (10, 20), (16, 19), (5, 11)] - your result is 19

The strange behavior of tests you see is because in this challenge, for some reason, tests always shuffle inputs before calling your function. This is why your solution sometimes passed, and sometimes fails: because tests sometimes generate inputs which trigger the bug in your solution, and sometimes not.