Infrared observations of NGC 3603 - III. The enigmatic, highly reddened sources of IRS 9

It is shown that the low-metallicity tail of the stellar metallicity distribution predicted by simple Outflow models for the Milky Way halo depends sensitively on whether instantaneous recycling is adopted or relaxed. In both cases, current - and still preliminary - data suggest a "G-dwarf problem" for the halo (reminiscent of the local disk). We suggest that the problem can be solved by introducing a (physically motivated) early infall phase. We point out several important implications of such a modification, concerning: the putative Pop. III (super)massive stars, the number of stars expected at very low metallicities, the questions of primary nitrogen and of the dispersion in abundance ratios of halo stars.