Investigating cell cultures with NMR requires high cell densities to provide adequate signal-to-noise, or scans must be summed over long time periods and short-term events are lost. The mixing within a chemostat can be used to shorten the time required to acquire informative in situ NMR spectra from cell cultures. However, performance trade-offs can occur between net signal, spectral resolution, and oxygenation due to sampling volume, conductivity, gas bubble, and fluid flow effects. These trade-offs and the effect of different mixing regimes were theoretically analyzed to quantify how device design decisions impact performance. The results were found to concur with data from cell-free NMR experiments performed in 18 mS/cm conductivity medium. The results also guided the redesign of an NMR bioreactor in terms of relative radio frequency (rf) coil and sample dimensions and other factors. The design, which entails using chemostat mixing to shunt sample through a rf coil in ca. 0.4 s, provides adequate oxygenation for the 4-16% (v/v) cell suspensions examined. Gains realized include lower conductive losses, better magnetic field homogeneity, and the exclusion of gas bubbles from the sampling zone. These gains enable the acquistion of spectra from dilute (3-4% v/v) Saccharomyces cerevisiae chemostat cultures in 6.9 min with high resolution in both the orthophosphate and the beta-NTP regions. Samples with 16% (v/v) cells also yield useful spectra within 0.5-1.0 min.