The impact of midhaul-induced latency on the performance of a wireless communication system must be understood in order to be appropriately accounted for. To this end, we study the impact of midhaul delay and jitter on the scheduling of a cloud radio access network. A soft-real-time commodity hardware testbed is used to gather data using up to 16 virtualized distributed units. The empirical data is used to develop models to analyze the midhaul latency's impact on spectral efficiency and outage probability. Modeling takes into account the burstiness of deadline misses in order to replicate outage duration in addition to frequency. The recorded data is then used to derive an estimate of system scaling in terms of shortest transmission time interval supported for a set number of distributed units for a given deadline miss rate. Results show that scaling behavior depends on the target rate. More stringent targets emphasize the impact of tail latencies. The testbed suffers less than 0.1 % spectral efficiency loss from deadline misses at a transmission time interval duration of 500 µs with 16 distributed units.