Module::Used, Module::ScanDeps, Module::Extract::Use et al., Devel::TraceUse can all be useful in listing what modules you instructed perl to use (as opposed to what modules are "usefully used" which is what you want).

I assume the difficult bit will be to find a way to remove each of these modules from source in order to see the effect on compilation (as you suggested in a comment below). Removing from source code (e.g. with regex) can be trickier than finding them. Because finding them usually involves reading the %INC and I am not sure whether removing from %INC will have any effect at all.

One thing that comes to my mind, totally untested, is: instead of removing a module from source code and run, just prepend a dir in @INC which will contain "fake" module files (with empty content, bar the boilerplate). For example:

mkdir -p fakemodules/Data
echo "package Data::Dumper;1;" > fakemodules/Data/Dumper.pm
# this will fail:
perl -Ifakemodules -MData::Dumper -e'print Data::Dumper::Dumper([1])'
a=$?
# this succeeds:
perl -MData::Dumper -e'print Data::Dumper::Dumper([1])'
b=$?
if [ $a -ne 0 ] && [ $b -eq 0 ]; then echo "module is required"; fi
[download]

1 minute edit: I think choroba mentioned a similar technique for mocking modules and probably this is where I got it from

bw, bliako