A very common data paradigm is to have destructive loads of your data. This means that you want simply to copy the data source exactly and wipe out anything in the target table. Today, Gainsight really only supports incremental loads with updates and upserts from what I can see. What this does is create extra work for some data sets to flag them (e.g. soft deletes) or in some how indicate that they should not be considered in reporting or anything. So this adds extra work for the admin to design a process to flag these invalid rows of data, but it also can require extra work for admins or others to ensure this data is filtered out.
While we understand that supporting this type of thing can be considered risky, but that should not block this type of functionality being available. There are lots of risky things an admin could do (e.g. delete the entire table). This type of functionality should have visible warnings and such when configuring it and maybe even permissions associated to it, but it would greatly help with data integration.