For viewful timeseries collections, the unpack stage was prepended during view resolution. This meant that the pipeline with the unpack stage was parsed and validated again. For viewless timeseries we prepend the unpack stage after the pipeline was validated. Therefore, we lose a call to validate. 

One edge case that would've caught this is if the user pipeline is at the internal pipeline length limit and therefore adding the unpack stage would push the pipeline to over the limit.