I doubt that removing newlines would increase the speed in any noticeable fashion.
That is, the performance issues are likely not caused by line count but rather the size/number/type/cost of the elements/operations after the parsing.
The actual lexer that handles the newlines should see them no differently in the stream than any other character. Depending on exactly what context "source" means newlines have some effect semantically at the parser:
- CSS: none (ignoring embedded newlines)
- HTML: possible new text-nodes or different content
- JavaScript: automatic semicolon insertion (or embedded newlines)
However, there is no reason not to try a minified version quickly to see if it makes a difference and, more importantly, to satisfy your curiosity ;-) I would definitely heed the other suggestions as well, such as checking the page (everything) for validity.
Happy coding.