Garnett (1989) discusses the production of sulfur, and concludes that it should come predominantly from stars with masses larger than 10 solar masses. Thus, sulfur and oxygen are both produced in massive stars, and one might expect the abundance of sulfur to follow that of oxygen quite closely (i.e., there is unlikely to be a significant "time delay" problem between the two). Garnett points out that potential variations between S and O would point to variations in the IMF and that observed S/O ratios may be useful in constraining nucleosynthesis calculations.
Woosley & Weaver (1995) have produced an extensive set of nucleosynthesis calculations for massive stars varying both stellar mass and metallicity. The oxygen yields show little dependence on metallicity, but the sulfur yields do change. However, the important factor for sulfur yields is the treatment of the hydrodynamics, specifically the degree of "fall back". With changes in the hydrodynamics it is possible for the sulfur production to change dramatically from all of the newly produced sulfur being ejected to all of it falling back into the remnant.
Overall, Garnett found S/O measurements from dwarf galaxies averaged log (S/O) = -1.55 ± 0.1 which is close to the solar value of log (S/O) = -1.66 (Anders & Grevesse 1989), and no evidence for a systematic variation in S/O with O/H. However, he cautioned that more observations at higher and lower O/H were needed. Observations of the very low abundance dwarfs IZw18 and UGC 4483 also found nearly solar ratios of S/O (Skillman & Kennicutt 1993; Skillman et al. 1994), and Thuan et al. (1995) found a very narrow range in S/O for a sample of blue compact galaxies, consistent with that found by Garnett, and slightly higher than the solar value. From this observational evidence, it would appear that either the IMFs are very similar in dwarf galaxies or that sulfur and oxygen production in massive stars have very similar mass and metallicity dependences (or both).