Question: I would like to pick up on a recent question by Teep

In a recent question (a few hours ago) Teep failed to compute quantiles of a negative binomial random variable as soon as the probability of success is small (even for a value of 0.05 the computational time is very prohibitive).
Teep used the package Statistics:

I discovered inadvertently that the package Student[Statistics] worked perfectly on the same problem, even for a success probability as small as 3e-10

What is the necessity to have two packages that share theoritically some features especially if one of them (Statistics) contains a lot of flaws and is not capable to return an answer as the other (Student[Statistics]) does?
In the present case of numerically computing quantiles, why people decide to implement a good algorithm in one package and a poor one in the other? Or why they didn't verify if the two packages gave identical answers on the same problem?