Categoryorder not working when specifying plot type

When using the category order parameter, no plot is generated if the plot type is specified in the plot_ly call:

p <- plotly::plot_ly(type = 'bar')
colNames <- names(iris)
colNames <- colNames[-which(colNames == 'Species')]

for(trace in colNames){
  p <- p %>% plotly::add_trace(data= iris, x = ~ Species, y = as.formula(paste0("~`", trace, "`")), name = trace)
}

p  %>% layout(xaxis = list(categoryorder='total descending'))

With the following warning, which I don't understand since I'm not combining discrete and non-discrete data in the same axis:

Warning message:
Can't display both discrete & non-discrete data on same axis 

If I specify the type in add_trace instead of the plot, then it works correctly and the warning is gone:

p <- plotly::plot_ly()
colNames <- names(df)
colNames <- colNames[-which(colNames == 'Species')]

for(trace in colNames){
  p <- p %>% plotly::add_trace(type = 'bar', data= iris, x = ~ Species, y = as.formula(paste0("~`", trace, "`")), name = trace)
}

p %>% layout(xaxis = list(categoryorder='total descending'))

Maybe this is the intended behavior? If so, why does specifying the type in the traces works and not in the plot? The warning seems unclear to me.

Start by running the supplied R reproduction with iris and compare plot_ly(type = 'bar') against add_trace(type = 'bar'). Trace how plot_ly and add_trace propagate the type and how layout(xaxis = list(categoryorder = 'total descending')) handles the resulting traces. Done means the reproduction no longer emits the warning and category ordering works consistently.