@David_Burnham has said about all.
In addition to those, TH Weibull and other TH distributions have been better fits than their regular counterparts for some data. But their usefulness usually appear when there is a positive threshold. If it is because that you data have zeros, it is worth understanding why they are zeros.
Input errors, rounding? Those are data issues.
Failure on arrival? Then consider ZI distributions as David said. Or maybe delete them, because those are likely to be a separate issue from the reliability problem, assuming you have a reliability problem.