Neural audio codecs (NACs) enable efficient audio transmission but are often optimized for monophonic signals neglecting multi-channel scenarios. This limitation is critical in applications such as distributed microphone arrays or binaural hearables, where spatial processing after transmission of compressed signals can be important. In this work, we evaluate a range of NACs with respect to spatial cue preservation across two datasets. Results reveal that individual codecs perform differently, that preservation of spatial cues at low frequencies improves with bitrate, and that degradation is consistent at high frequencies. Although these effects may be perceptually subtle, they can significantly degrade the performance of downstream processing algorithms. We provide actionable guidance for choosing a NAC to transmit one or more channels before spatial processing.