This is a crucial step, in which all RFI affected or otherwise corrupted data is flagged. The three major sources of data corruption are (1) RFI, (2) loss of signal from an antenna (referred to as 'dead antenna'), and (3) correlator problem.
The GMRT correlator naturally generates the time-frequency plane data for each baseline with up to 256 frequency channels . As shown in the work on auto-flagging (reference 3), this can be effectively used to remove outlier visibility points by examining the time-frequency plane data from cross- and self-correlations. Except a threshold setting, this flagger is reasonably automatic and robust. A program for this already exists and is in use on the GMRT data by various users of the telescope. Note that the threshold setting is easiest for an unresolved or moderately resolved source. Since all the sources we will use will be in this category, it is reasonable to assume that a large amount of flagging of type (1) and (2) above can be automated.
Experience shows that it is possible to identify data corrupted due to
correlator problems and/or inordinately high 
at a few
antennas using the robust algorithm (see reference 2). Again, this
works best for unresolved or moderately resolved sources, which will
be the case for all our sources.
It appears therefore that it is possible to implement a reasonably automated procedure for flagging corrupted data.