Why can't AalenJohansenFitter handle multiple failure events occurring at the same time?

[anthonymichaelclark]

Coming from here

#619 (comment)

I've been learning competing risks modeling out of the book "Competing Risks and Multistate Models with R" (and promptly trying to find Python solutions since that's what the rest of our stack runs in :) ). The authors propose the following non-parametric estimator for the probability of failure due to cause j prior to time T:

They call this estimator a "special case of the Aalen-Johansen estimator for general time-inhomogeneous Markov processes".

It doesn't look to me like this estimator has any trouble with failure events occurring at the same time, so why does AalenJohansenFitter throw warnings//perturb failure time data to fit? Does the general case of an Aalen-Johansen estimator preclude it somehow and if so, is there a term for that^ estimator that's more descriptive? Would maintainers be interested in an implementation for that^ estimator, possibly one that fits all competing events simultaneously?