Custom dimensions for a ggplot2 boxplot doesn't work with ggplotly

When I create a ggplot2 boxplot with custom dimensions (see the bottom of http://ggplot2.tidyverse.org/reference/geom_boxplot.html#examples), plotly doesn't render the boxplot when I apply ggplotly. However, it works ok with a traditional boxplot where the dimensions are calculated by geom_boxplot/stat_boxplot.

Below is a reproducible example of the issue.

## Libraries
library(ggplot2)
library(plotly)

## Using fake data with extreme values
fake_data = data.frame(
  y1 = c(rnorm(100, 2, 3), rnorm(10, 10, 15))
)


# ggplot2
g1 = ggplot(data=fake_data, aes(x ="cat1", y= y1)) + 
  #stat_boxplot(geom="errorbar") +
  geom_boxplot() +
  labs(title="Traditional Boxplot") +
  ylim(c(-5, 25))

g1

quants = quantile(fake_data$y1, c(0.05, .25, .5, .75, .95))
q_df = data.frame(t(quants))
names(q_df) = c("y5", "y25", "y50", "y75", "y95")


g2 = ggplot(data=q_df, aes(x ="cat1")) + 
  #stat_boxplot(geom="errorbar") +
  geom_boxplot(stat="identity", aes(, 
       ymin = y5, lower = y25, 
       middle = y50, upper = y75,
       ymax = y95)) +
  geom_point(data=subset(fake_data, (y1 > q_df$y95) | (y1 < q_df$y5)), aes(y=y1)) +
  labs(title="5%/95% Boxplot") +
  ylim(c(-5, 25))

g2



# plotly
plotly::ggplotly(g1)
plotly::ggplotly(g2) # DOESN'T WORK!

Start with the reproducible R example in the issue and compare ggplotly(g1) with ggplotly(g2), focusing on the stat="identity" boxplot and custom ymin, lower, middle, upper, and ymax mappings. Reproduce the failure, then trace the ggplotly conversion path for these boxplot values; done means the custom-dimension plot renders while preserving the outlier points.