[racket] How to tell if identifier is bound in racket/load?

Hi,

I've been using the following awesome snippet to detect when
identifiers are already bound:

; From Danny Yoo's arctangent
; https://github.com/dyoo/arctangent/blob/master/language.rkt
; Returns true if stx is an identifier that is lexically bound.
(define-for-syntax (lexically-bound? stx)
  (let ([expanded (local-expand stx (syntax-local-context) #f)])
    (cond
      [(and (identifier? expanded)
            (eq? #f (identifier-binding expanded)))
       #f]
      [else
       #t])))

Then I created a helper so I could call it at runtime:

(define-syntax (arc-bound stx)
  (syntax-parse
   stx
   [(_ var:id)
    (if (lexically-bound? #'var) #'#t #'#f)]))

That still worked great.  Then I tried using racket/load:

Welcome to DrRacket, version 5.2.1 [3m].
Language: racket/load; memory limit: 128 MB.
> (arc-bound a)
#f
> (define a 5)
> (arc-bound a)
#f
>

:(  I've read through the docs on namespaces and the racket/load
language source in an attempt to fully understand why racket/load is
different, but I'm missing something.  I've noticed that I can see 'a
in (namespace-mapped-symbols), and have tried using that, but ran into
(I think) phase issues.  If there's a quick solution to this that I've
missed, that's brilliant.

Otherwise, my motivation to start making use of racket/load (really
arc/load in my case), is to get access to global variables.  If
ark.rkt declares something at the module level, I need to be able to
see it - the same copy of it - from other modules that require
ark.rkt.  What I'm trying now is to have ark.rkt be a real module
written in arc, and then require it and load the rest of the arc files
using the arc/load language so they all share the same module registry
and the same ark.rkt.  If there's some other, better way to do this,
pointers will be quite warmly received.

I'm also becoming aware that at least for now I'm probably one of
those users who knows just enough to be dangerous, but not enough to
save myself.  I welcome challenges to my approach, corrections to my
descriptions, and inquiries for more information.  It's possible I'm
asking for details about the completely wrong approach, when really
something else would be more appropriate.

-Nick