Recommended way to handle RTL mixed with LTR languages

[Mittineague]

I’m wondering if the bottleneck is Discourse generating the directional instructions or the browser interpreting them and rendering the display.

Have you compared the efficiency using different browsers?