This is related to issues 06429, 06381 and 06506, but this one concerns the mapping of attributes.

When mapping participant attributes to token attributes (in the process of adding a participant to a survey), resulting token attribute descriptions are not filled if admin language is different from the participant attribute language.

If admin language and participant attribute language are the same (let's say english), then the resulting token attribute description is correctly populated with the attribute name.
For example, if original attribute name (in english) is 'Age', then the resulting token attribute will be called attribute_cpdb_xxxx and its description will be set to 'Age' (and 'Age' will appear as a column in the tokens list).

If, on the other hand, admin language is french, then when mapping the same attribute to a survey, the resulting token attribute will also be called attribute_cpdb_xxxx but the description will be empty (instead of using 'Age'). And as another result this token attribute will be displayed as attribute_cpdb_xxxx in the tokens list instead of 'Age'.

I think the right way to do this is to use the name of the attribute regardless of admin language (which is obviously currently used to determine the attribute description to use in the process). If the original attribute is multi-lingual (has n names for n languages), then the target survey language should be used to choose the right language description to copy over.