Answer to this question -- https://twitter.com/crude2refined/status/852198071727468544

dynamic-y.R

library(plotly)
library(tidyr)
library(crosstalk)

m <- gather(mtcars, variable, value, -vs)
msd <- SharedData$new(m, ~variable)
gg <- ggplot(msd, aes(factor(vs), value)) + 
    geom_jitter(alpha = 0.3)

bscols(widths = c(11, 6, 6),
    filter_select("id", "Select a variable", msd, ~variable, multiple = FALSE),
    ggplotly(gg, dynamicTicks = TRUE) %>% layout(margin = list(l = 30)),
    plot_ly(msd, x = ~jitter(vs), y = ~value) %>% add_markers(alpha = 0.3)
)

Sorry, I didn't make this example with the intention of doing (1).

If I did, I would have done, msd <- SharedData$new(m) and would have added highlight(off = "plotly_deselect"), which would solve (2) and (3). The default "off event" is a relayout which can be triggered via the home button

I'm aware of (4), and partially why I made the example, plotly/plotly.R#944

Thanks for bringing up (5), I've seen this before too, but I think it's something that needs to be addressed in plotly.js

I just wanted to show the two different ways of acheiving jitter and acheviing this