Shaded jars are useful as an end artifact, but are problematic to reuse as a dependency. So we should publish both shaded and unshaded artifacts.
Currently for some modules in analytics/refinery expose shaded jars as the main artifact. We can either publish an unshaded jar as the main artifact and publish the shaded one with a classifier, or the opposite.
The usual expectation is that the main artifact from a module will be a standard jar. A shaded jar make sense as a main artifact when the module is a pure entry point (just a main class, only code that make no sense to reuse). It is a good practice to have a dedicated module for an entry point (which might or might not make sense in this project).
In case like this project, where entry points and supporting classes are in the same module, the usual is to have the standard jar as the main artifact, and publish a shaded jar with a classifier (which is what I implemented).
It is possible to publish the shaded jar without classifier and publish the unshaded jar with a classifier. This would break the principle of least surprise (at least for people familiar with maven). It would also require quite a bit more configuration.
So if adding a classifier to the shaded jar is an issue (lot of surrounding code to update), I can try to implement something else.