At FMX, they measuring rocking curves in a Jupyter notebook and refer back to the notebook as a logbook of what was done. They do not want any over-plotting. They want something like the user experience of %matplotlib inline, but with a live-updating plot during acquisition. It also wouldn't hurt to retain interactivity (i.e. pan and zoon) after the scan, but this is not essential.

To address this, would could have a JupyterFigures variant that places each new figure in the currently-executing cell (by calling display(...) rather than adding a tab to an exiting widget. I am not 100% sure this is technically feasible, so we'll need to try it.