Rosetta Code would be a good starting point (depending on your personal definition of "easy" and "hard", of course).

Agreed. I find that the best place most of the time to find learning/sample code is the test suites of Perl packages. Perl has an above average testing culture I think. This turned up a few interesting things and I’m sure there is plenty more in there–

moo@cow[63]~/.rakudobrew>ack \\bloop\\b -l moar-v6.c/ | ack \\.t$
moar-v6.c/nqp/examples/rubyish/t/loops.t
moar-v6.c/nqp/examples/rubyish/t/template.t
moar-v6.c/nqp/t/jvm/01-continuations.t
moar-v6.c/nqp/t/moar/01-continuations.t
moar-v6.c/nqp/t/moar/02-qast-references.t
moar-v6.c/nqp/t/nqp/14-while.t
moar-v6.c/nqp/t/nqp/15-list.t
moar-v6.c/nqp/t/nqp/42-cond-loop.t
moar-v6.c/nqp/t/nqp/84-loop-labels.t
moar-v6.c/nqp/t/parrot-pmc/qrpa.t
moar-v6.c/nqp/t/qast/01-qast.t
[download]

At my day job, there is only tests for approximately 20% of all code. They've finally mandated that all files get at least 80% coverage (this is all Python). I'm a stickler in my Perl apps to reach at least 90% coverage, but frequently reach 95+, testing most things in numerous way at that.

I'm also currently half way done a tutorial for PerlMonks that shows the whole github, travis and coveralls configuration so those who don't know, will easily be able to implement and appreciate CI