Density plot draws unwanted outlines (similar to previously-removed release of ggplot2)

A density plot created with ggplot2 which has no outline then has an outline drawn when used with ggplotly. Relevant changelog for ggplot2.

Simple version of minimal reprex:

mtcars <- mtcars

p <- ggplot(mtcars, aes(x=disp)) + geom_density()
ggplotly(p)``

A workaround for this simple example (proposed by MrFlick on stackoverflow) was to use stat_density :

p <- ggplot(mtcars, aes(x=disp)) + stat_density(geom="line")
ggplotly(p)``

However, my 'real' data has multiple groups with different distributions. In this case, ggplotly still draw the unwanted outline, but also cannot be fixed this time by stat_density. Example:

p <- ggplot(mtcars, aes(x=disp, group=carb, color=carb)) + geom_density()
ggplotly(p)

p2 <- ggplot(mtcars, aes(x=disp, group=carb, color=carb)) + stat_density(geom="line")
ggplotly(p2)

Empieza ejecutando el reprex mínimo con ggplot2 y ggplotly; después, compara los ejemplos de densidad de un solo grupo y de grupos con la solución alternativa stat_density. Se considera terminado cuando ggplotly ya no dibuja contornos no deseados para los gráficos de densidad, incluidas las distribuciones agrupadas.