I think we can distinguish 3 different problems here:

1. Natural language spellchecking. That's what this issue is about.

2. Spell-checking code. (Essentially, identifiers.)

3. Finding code inside natural language, and checking it as if it were code. (That's what you're talking about here.) This is not a spellchecking problem, it's a problem of identifying which spell-checking apparatus to use, rather like font-lock for multi-language buffers. It's hard to see how to do it without some syntactic clue (e.g. the use of backticks in markdown), as used in multi-language buffers for font-locking.