In some bidi locales, combinations of number symbols & numbers patters, as well as short date formats, have been carefully curated to work when they have both a strong RL context and a neutral or default context. However, the Survey Tool only shows examples with a strong RL context, which may lead vetters to simplify patterns or strip marks so the patterns only work well in a strong RL context.
For example, as a result of CLDR 32 ST changes, fa/arabext number patterns/symbols changed as follows (shown in ICU format):
suggest 36-optional, major; maybe for Tom or Peter?
In ExampleGeneratorl; current code may generate one bidi context if text contains any RL chars, but could easily generate two bidi contexts
Would you be able to review and see if this can be resolved in v38?