Because the ATCA has only a modest number of baselines, it will often be essential that you combine data that come from a different time and configuration. This is done to improve the uv coverage of the observation. In principle, you could combine multi-source data bases with DBCON, and calibrate the concatenated data base. However, for a number of reasons, I think this approach is more trouble than it is worth, and I suggest that the data for each configuration should be calibrated (ATCALIB, ATCLCAL etc.) separately, split (SPLIT) off into a single-source file, and then combined (DBCON). The imaging tasks can then be run on the combined data base (see § 15). I do not recommend concatenation of multi-source files after they have been calibrated, as the calibration tables are not always correctly merged. I also suggest that you combine TB-sorted data, as this should be the order of your multi-source file. It doesn't matter what time order the different configuration single-source files are concatenated in.
Recall that SPLIT applies the calibration directly to the visibilities and builds single-source files for one source and frequency. The only calibration table that might become associated with the single-source file in the future is the SN (solution) table, because new ones will be created in the self-calibration procedure by CALIB (see § 17).