In the manual for the Refinement fraction description it says " The fraction of cells with the largest error that should be flagged for refinement." which kind of reads like a value of 0.25 will refine 1/4 of the cells, and the cells that are chosen for refinement will be the ones with the largest error. However, the description for the parameter Adapt by fraction of cells (which is not the default) says " 0. Use fraction of the total number of cells instead of fraction of the total error as the limit for refinement and coarsening." So this seems to suggest that the default process is not to take a fraction of the cells, but rather the cells that account for a fraction of the error. So, this would mean that the default behavior is to determine the total error, and then for a refinement fraction of 0.25, choose to refine the cells that account for top 25% of the total error. (When I say top, I’m imaging of cells sorted by error with the cells with the largest error at the top). Similarly, a Coarsening fraction of 0.10 would coarsen the cells that account for the bottom 10% of the error. Am I interpreting this correctly, or am I making things more complicated than they really are?
Yes, you’re interpreting this correctly!
(And if it’s not clear, we should probably improve the documentation in the manual…)
Yeah - thank you! It is a bit fuzzy in the manual. Maybe i can do this a warm up pull request for the hackathon.
I’ve had that same question in the past so I think clarifying that in the manual would be great - and yes, it seems like it would also be a nice warm up for the hackathon 
Let me know if I can help.