Extragalactic H II regions, direct tracers of luminous young stars, are traditionally studied in H, H, and other lines characteristic of gas at K. Radio observations, though difficult, offer two key advantages.
A single O5 star creates an H II region with a flux density of about 1 mJy at 1 Mpc, or 10 microJy at 10 Mpc; a B0 star is 100 times less bright. The enhanced VLA at 40 GHz will reach a sensitivity of 2 microJy in 12 hrs with 1 GHz bandwidth; the super-wide 10 GHz bandwidth would make this much easier (1.3 hrs), but in any case we could count individual stars down to the B0 class in nearby galaxies. Even with the narrow 1 GHz band we could see a single O5 star at 10 Mpc in 12 hrs, and 10 GHz would allow either very fast imaging of distant sources, or a more involved but still plausible project to look at even smaller H II regions out to Virgo. One should pick up many more of the optically-detected H II regions, allowing the determination of extinctions, the direct measurement of temperatures and densities (useful for example in abundance work), and high-resolution imaging (see below). This might also be a good approach to finding possible star-forming regions in dwarf, elliptical, and low (optical) surface brightness galaxies, although direct H imaging probably wins when looking simply for detections.
Long-wavelength observations would be useful in distinguishing H II regions from faint background sources and SNRs, requiring a sensitive long-wavelength system (somewhere between 0.6 and 2.4 GHz). One would of course also get SNRs for free (see above), making this a wonderful star formation project.
The high angular resolution of the A+ configuration would be especially useful for this work. A typical H II region is a few to tens of parsecs across, and can be resolved in nearby galaxies even with the current VLA. Compact (pc) and ultra-compact (pc) H II regions, corresponding to much younger sources, are much more difficult to study outside our own galaxy. At 9mm the A configuration gives resolution, or 0.3 pc at 1 Mpc, and 3 pc at 10 Mpc; the extended A+ configuration is a factor 7 better, or 8 mas (0.04 pc at 1 Mpc, 0.4 pc at 10 Mpc, 0.55 pc at Virgo). Given the sensitivities discussed above, one could easily detect and image hundreds of individual ultra-compact H II regions in nearby galaxies, and even resolve compact H II regions to the distance of Virgo. This means knowing the ages and (high end) initial mass functions for star forming regions across entire galactic disks, in unprecedented detail, with no worries about extinction. This simply cannot be done now. (The MMA would do about as well, although in both cases, to eliminate confusing sources, the expanded VLA with high-sensitivity low-frequency systems would be needed.)