I have a bokeh page of 40 tabs, each with 6 time series plots of length
~2000. That page is over 50mb (less an issue) and takes over 30 sec to
load. Chrome even thinks the page is not responsive and asks if I should
wait or kill. (Click wait and the page eventually shows)
Does it help if I use less tools? Is there other trick to make it load
quicker?
No tool can be ideal for every possible use case. Bokeh makes explicit trade-offs in some areas, in the name of supporting very sophisticated interactivity. In very rough terms, Bokeh performs much better on the "more points but fewer plots" side of the scale. This is because every plot has to go through a fairly sophisticated and involved initialization and setup to support Bokeh's interactive features. It's possible that some of this setup (such as building spatial indices for hit-testing) could be made optional, but that would represent new development and is not currently configurable.
If you do need the highly interactive features of bokeh (linked behaviors, tooltips, customjs or bokeh server callbacks, etc.), then best thing you can probably do is to break your page up into multiple sub-pages somehow. If you aren't making use of Bokeh's interactive features, then I might suggest that you'd be better off with a different tool that generates static images for plots (e.g. datashader, or matplotlib, etc.)
I have a bokeh page of 40 tabs, each with 6 time series plots of length ~2000. That page is over 50mb (less an issue) and takes over 30 sec to load. Chrome even thinks the page is not responsive and asks if I should wait or kill. (Click wait and the page eventually shows)
Does it help if I use less tools? Is there other trick to make it load quicker?