Opacity in Grouped Bar Charts with data.frame

When plotting a grouped bar chart with variable opacity with data in a data.frame, the opacity assigned to each column is not the one specified in the data.frame. In the example below, the 2 bars in group "A" are plotted with reduced opacity, as opposed to the 2 bars with colour "C". Strangely, if the data.frame is not ordered then it plots correctly.

library( plotly )

data  = data.frame( 
  x      = c( "A", "B", "A", "B" ),
  y      = c(   1,   2,   3,   4 ),
  z      = c( "C", "C", "D", "D" ),
 opacity = c( 0.3, 0.3,   1,   1 )
)

# removing this lines the plot is correct
data = data[ order( data$x ),]

p = plot_ly( 
  data   = data, 
  x      = ~x, 
  y      = ~y, 
  color  = ~z, 
  marker = list( opacity = ~opacity ),
  type   = "bar" ) %>% 
layout( barmode = "group" )
show( p )

Start by running the supplied R reproduction in plotly.R with the data frame both ordered and unordered. Trace the grouped bar chart handling of marker opacity and compare each bar's opacity with the data frame values. Done means ordering the data no longer changes which bars receive reduced opacity.