There might be a situation when a language's token sequence is divided to several pieces by tokens of higher level
language. For an example see issue #111546. 

The CSS code embedded in HTML which is embedded in JSP might contain some jsp tokens in basically two ways:
1) the embedded tokens replaces/generates a part of the css code ( h1 { color: ${"red"}; } )
2) the embedded tokens DOESN't replace/generate part of the css code ( h1 { color: red; ${"   "} } )

case #2 can work even without 'fix' of this issue, however case #1 cannot AFAIK solved in a generic way. In such case we
need to adjust particuallar rules in the grammar to be able to properly reflect the embedded tokens. 

With the gap terminal we can for example modify the grammar rule for css rule item following way:

RULE_ITEM -> NAME : (VALUE | gap) ;

so if user types:

<style>
 h1 { color: ${"red"}; }
</style>

the grammar is able to parse and the features will work, at least I think ;-)