Unexpected behavior with geom_rect and ids arg in ggplotly()

Expected behavior demonstrated by geom_point. See ggplotly(gg1) below.

Unexpected behavior when using geom_rect with the ids arg across frames with changing number of observations. See ggplotly(gg2) and ggplotly(gg3) below.

### Win 10 x64, Rstudio
library(ggplot2) # ggplot2_3.0.0 
library(plotly)  # plotly_4.7.1
data(gapminder, package = "gapminder")

# filtering out some countries (soframes have differing number of obs)
sub_countries <- unique(gapminder$country)[1:10]
gapminder2 <- gapminder %>% 
  filter(((year != 1952) | (country %in% sub_countries)) & year < 1970)

# using geom_point (works)
gg1 <- ggplot(gapminder2) +
  geom_point(aes(x = gdpPercap, y = lifeExp, 
                 ids = country,
                 frame = year)) +
  scale_x_log10()
ggplotly(gg1) # correct: sub, full, full, full

# using geom_rect (doesn't work with ids arg)
gg2 <- ggplot(gapminder2) +
  geom_rect(aes(xmin = gdpPercap, ymin = lifeExp, 
                xmax = 1.05 * gdpPercap, ymax = 1.05 * lifeExp, 
                ids = country,
                frame = year))  +
  scale_x_log10()
ggplotly(gg2) # incorrect: sub, sub, sub, sub

# using geom_rect, subset on 2nd frame (doesn't work with ids arg)
gapminder3 <- gapminder %>% 
  filter(((year != 1957) | (country %in% sub_countries)) & year < 1970)

gg3 <- ggplot(gapminder3) +
  geom_rect(aes(xmin = gdpPercap, ymin = lifeExp, 
                xmax = 1.05 * gdpPercap, ymax = 1.05 * lifeExp, 
                ids = country,
                frame = year))  +
  scale_x_log10()
ggplotly(gg3) # incorrect: full, sub moves, full, full
  # frame year 1952 -> 1957 != 1962 -> 1957 != 1967 -> 1957

Cheers,
Nick

Ejecuta la reproducción de R proporcionada con ggplot2 y plotly, comparando geom_point y geom_rect entre frames con un número cambiante de observaciones. Traza la ruta de conversión de ggplotly para geom_rect, ids y el manejo de frames. Se considera terminado cuando geom_rect conserva las observaciones y los ids correctos en cada frame, coincidiendo con el comportamiento esperado demostrado.