When it is quite difficult to wrap around those existing GUI implementation, another way to look at the issue is that, we may need a standard for Perl GUI.
The standard should specify what wighets are mandatory for all Perl GUI implementation, what are optional, and for each wighet, what its default rendering is, what its default behavior is, what API's it should support...
Tk should be repackaged. Tk does not provide us enough functionality we want, but at the same time, Tk is a little bit fat, if you take a look at how many modules are bundled under Tk. Also some of the high-end widgets are poorly developed. Actually if Perl provides a strong core GUI implementation (a kind of leadership), it might attract the majority of Perl developers, and become the de facto standard, which is a role Tk failed to play so far.
For example, Tk::Table is one of those poorly designed widget. It is not easy to use (alomst cannot be used directly), and its default behavior (provided) is really limited. One might view this as flexibility, but I would argue, fair enough, but then let's have a sub class of Tk::Table, for example Tk::Table::Default, to provide a concret default table implementation, and bundle it as core module.