(posting answer here as well in addition to email conversation)
Some code in my URL encoding routine (which works by splitting parameters and encoding them) is what converts
http://www.example.com/example.cgi/GLOS?;text;
to
http://www.example.com/example.cgi/GLOS?text;
I am not 100% sure if A1SG is wrong about "fixing" the URL and will need refresh my memory, but e.g.
http://www.example.com/example.cgi/GLOS?&text& would be an extremely uncommon URL as well. But possibly semicolons supports different usage than ampersand, I will need check some specs

That said, URL encoding code should not change the URL unnecessary, so I am thinking about reworking it and then add the "URL fix" into separate code and option.
you can find the URL encoding option in:
Scan website | Crawler options | URL encode query params