How if at all does Pandoc handle URL encoding/percent encoding?

[bpj]

(Apologies if this is old hat!)

1. Does Pandoc do any automatic URL/percent encoding in output URLs?
2. Does Pandoc decode in input URLs?
3. If (1) does Pandoc recognise when an URL already is wholly or partly encoded?
4. If (1) and (3) can I rely on URLs not becoming double-encoded?
5. What about identifiers/fragments? I'm not quite sure about the rules. (w) Is encoding at all supported in those, and (b) if so how does Pandoc handle it?

It seems like already encoded URLs are passed through as is,
but I want to be sure what gives. Somewhat heavy-handed
encoding¹ with Lua is easy, but is it even needed? If it
is what about including (smarter) URL
encoding/decoding functions in the Lua API?

Footnotes

1. -- os.setlocale('C')
   local url_encode
   do
     -- byte('!') --> %21
     local pc_hex_fmt = '%%%02X'
     local byte2pc_hex = function(b)
       return pc_hex_fmt:format(b:byte())
     end
     url_encode = function(url)
       -- TODO Return nil for nil?
       if url == nil then
         return ""
       end
       -- Must be a string (or nil)
       if not ('string' == type(url)) then
         error('Argument is not a string')
       end
       -- Return early if empty
       if "" == url then
         return ""
       end
       -- Network newlines
       url = url:gsub("\n", "\r\n")
       -- Encode all except unreserved and SPACE
       -- DONE More fine-grained selection of "unsafe" chars
       url = url:gsub('[^%w%-%_%~%.\x20]', byte2pc_hex)
       -- SPACE --> +
       url = url:gsub("\x20", '+')
       return url
     end
   end
   

   ↩