@jrbourbeau

This seems fine but I would also like to understand what the motivation for this is:

defaults.update((k, v) for (k, v) in kwargs.items() if k in _get_figure_keywords())

Ultimately the pro-active property checking of _get_figure_keywords is what tripped things up in this case, IIUC. Generally speaking this is not something we would encourage normally, so I'd be interested to see if we can offer alternatives that satisfy your intentions here.

Thanks for taking a look @bryevdv. From what I can tell the introduction of keyword argument validation with _get_figure_keywords stems from wanting to plot multiple profilers using dask/diagnostics/profile_visualize.py::visualize, which have similar but not exactly the same signatures for their corresponding plot_* utilities in dask/diagnostics/profile_visualize.py (e.g. plot_cache). That's all to say, I think this is an issue on the Dask side of things and not the bokeh side.

I'm going to merge this PR as it's a good step forward and will help resolve CI build failures. Though I agree revisiting this, and trying to eliminate the need for _get_figure_keywords, is something we should do