Recent investigation of the extinction law in 30 Dor and the Tarantula Nebula, at optical and near infrared (NIR) wavelengths, has revealed a ratio of total to selective extinction R_V=A_V/E(B-V) of about 4.5. This indicates a larger fraction of big grains than in the Galactic diffuse interstellar medium (ISM). Possible origins include coalescence of small grains, small grain growth, selective destruction of small grains, and fresh injection of big grains. From a study of the ultraviolet extinction properties of three massive stars in the 30 Dor nebula (R139, R140, R145), observed with the International Ultraviolet Explorer (IUE), we show that the excess of big grains does not come at the expense of small grains, which are still present and possibly even more abundant. Fresh injection of large grains appears the dominant mechanism. A process able to naturally account for this in environments such as the Tarantula nebula, where formation of massive stars has been ongoing for over ~20 Myr, is the explosion of massive stars as type-II supernovae (SN). The ensuing change in the conditions of the ISM is only temporary, lasting less than ~100 Myr, because shattering and shocks will eventually break and destroy the bigger grains. However, this is the only time when star-forming regions are detectable as such in starburst and high-redshift galaxies and we highlight the complexity inherent in interpreting observations of star-forming regions in these environments. If the extinction characteristics are not known properly, any attempts to derive quantitative physical parameters are bound to fail.