The identification of the first confirmed neutron star -black hole (NS-BH) binary mergers by the LIGO, Virgo and KAGRA collaboration provides the opportunity to investigate the observed properties of the early sample of confirmed and candidate events. Here, we focus primarily on the tilt angle of the black hole's spin relative to the orbital angular momentum vector of the binary, and the implications for understanding the physical processes that determine this tilt. While the tilt angle of GW200105 is unconstrained, the posterior tilt distributions of both GW200115 and the candidate event GW190426_152155 peak at significantly anti-aligned orientations (though both display wide distributions). If taken close to the peaks of their posteriors, both would be significant outliers from the posterior predictive tilt distribution of binary black hole mergers. Producing these tilts would require stronger natal kicks than are typically considered (and preferentially-polar kicks would be ruled out), and/or an additional source of tilt such as stable mass transfer. The early sample of NS-BH events are less massive than expected for classical formation channels, and may provide evidence for efficient mass transfer that results in the merger of more massive NS-BH binaries before their evolution to the compact phase is complete. We predict that future gravitational-wave detections of NS-BH events will continue to display total binary masses of ≈ 7 M and mass ratios of 𝑞 ∼ 3 if this is the correct interpretation. Large tilts in a significant fraction of merging NS-BH systems would weaken the prospects for electromagnetic detection. However, we show that EM observations, including non-detections, can significantly tighten the constraints on spin and mass ratio.