Strongly gravitationally lensed galaxies are veritable gems for our understanding of high-redshift galaxy evolution, allowing us (at least in the few most fortuitous cases) to study dust, gas, and star formation even in the most intense starbursts on scales of individual star-forming regions. On these scales, below one to few 100 pc in the source plane, kpc-scale rotational support no longer dominates, and star formation is regulated by the local gas and stellar mass surface densities and energy injection from turbulence and winds driven by star formation and AGN. In recent years, wide-field far-infrared and sub-millimeter surveys carried out with Herschel and the South Pole Telescope have drastically increased our samples of such starburst galaxies. The brightest such sources known, however, come from another, even larger survey, which was certainly not built to study high-redshift galaxies: The Planck all-sky survey. In spite of 5' beam size and a confusion limit of 600 mJy (corresponding to a FIR luminosity of about 5x10^13 Lsun at z~2), Planck has provided us with a small, exquisite set of the brightest, strongly gravitationally lensed, dusty, high-z starburst galaxies known, 'Planck's Dusty GEMS'. I will report on our follow-up work that was necessary to identify and characterize these galaxies, including ALMA interferometry at 0.1'' beam size, and discuss the first surprising lessons which they have already taught us, from the signatures of winds to diffuse cold neutral gas to the delicate balance between turbulence and gravity, which does not seem so different from that in giant molecular clouds in the Milky Way. I will also discuss the first direct observational constraint for a (bottom-heavy) stellar initial mass function from gravitational lensing in the galaxy-formation epoch, which is consistent with those of massive low-redshift early-type galaxies, but defies the standard assumptions of most cosmological models.