Hi Michel!
I kind of like your proposal. If nothing else, it is at least a nice
experiment.
There is one thing I didn't understand: In the introduction you talk
about only affecting "symbols explicitly declared to be 'lexical'",
while other parts of the proposal seem to suggest that the lexical
layer would once loaded always be in effect. Which way do you want
to go?
I think I would like the first possibility. I.e. something like
with_lexical([x], f(x):=x[1]);
Regards,
Harald