Jupiter's moon Io hosts a dynamic atmosphere that is continually stripped off and replenished through frost sublimation and volcanic outgassing. We observed an emission band at 1.707 µm thought to be produced by hot SO molecules directly ejected from a volcanic vent; the observations were made the NIRSPEC instrument on the Keck II telescope while Io was in eclipse by Jupiter on three nights in 2012-2016, and included two observations with 10× higher spectral resolution than all prior observations of this band. These high-resolution spectra permit more complex and realistic modeling, and reveal a contribution to the SO emission from gas reservoirs at both high and low rotational temperatures. The scenario preferred by de Pater et al. (2002) for the source of the SO gas -direct volcanic emission of SO in the excited state -is consistent with these two temperature components if the local gas density is high enough that rotational energy can be lost collisionally before the excited electronic state spontaneously decays. Under this scenario, the required bulk atmospheric gas density and surface pressure are n ∼ 10 11 cm −3 and 1-3 nbar, consistent with observations and modeling of Io's dayside atmosphere at altitudes below 10 km (Lellouch et al. 2007; Walker et al. 2010). These densities and pressures would be too high for the nightside density if the atmospheric density drops by an order of magnitude or more at night (as predicted by sublimation-supported models), but recent results have shown a drop in SO 2 gas density of only a factor of 5±2 (Tsang et al. 2016). While our observations taken immediately post-ingress and pre-egress (on different dates) prefer models with only a factor of 1.5 change in gas density, a factor of 5 change is still well within uncertainties. In addition, our derived gas densities are for the total bulk atmosphere, while Tsang et al. (2016) specifically measured SO 2 . The low-temperature gas component is warmer for observations in the first 20 minutes of eclipse (in Dec 2015) than after Io had been in shadow for 1.5 hours (in May 2016), suggesting cooling of the atmosphere during eclipse. However, individual spectra during the first ∼30 minutes of eclipse do not show a systematic cooling, indicating that such a cooling would have to take place on a longer timescale than the ∼10 minutes for cooling of the surface (Tsang et al. 2016). Excess emission is consistently observed at 1.69 µm, which cannot be matched by two-temperature gas models but can be matched by models that over-populate high rotational states. However, a detailed assessment of disequilibrium conditions will require high-resolution spectra that cover both the center of the band and the wing at 1.69 µm. Finally, a comparison of the total band strengths observed across eight dates from 1999-2016 reveals no significant dependence on thermal hot spot activity (including Loki Patera), on the time since Io has been in shadow, nor on the phase of Io's orbit at the time of observation.