layout breaks with facet_wrap when employing scales=free

I have a problem combining plotly::ggplotly (v4.7.1) with facet_wrap (v3.0.0) that I can't seem to generalise but is reproducible with a particular dataset (summary metrics for a set of tweets):

require(tidyverse)
require(plotly)
d = read_csv('https://gist.githubusercontent.com/geotheory/21c4eacbf38ed397f7cf984f8d92e931/raw/9148df79326f53a66a8cc363241a440752487357/data.csv')
d = d %>% mutate(key = fct_reorder(key, n)) # order the bars
p = ggplot(d, aes(key, n)) + geom_bar(stat='identity') + 
  facet_wrap(~ set, scales='free', nrow=1) +
  labs(x=NULL, y=NULL) + coord_flip()
print(p)

Enter ggplotly:

print(ggplotly(p))

This seems to relate to the combination of nrow=1 and scales='free' arguments. Any ideas about the cause?

首先，使用链接的数据集和报告中的 ggplot2 与 plotly 版本运行所提供的 R 示例，并比较 ggplot 和 ggplotly 的渲染结果。跟踪 ggplotly 如何处理带有 scales='free'、coord_flip() 和 nrow=1 的 facet_wrap。完成的标准是：转换后的图保留预期的 facet 布局，同时不破坏自由 scale。