Background
The pipeline will occasionally produce a long solution interval on the order of 1000s. This happens when an 8bit setup is used two or more times, separated by a different setup. The sol int will span from the last phase scan of the first group to the first phase scan of the next group of same setup. This is related to combine='scan' in gaincal. In the worst cases, an additional effect seems to be that the pipeline somehow averages data between different fields.