Hi,
from your description I am assuming the data inside your source's tables will be deleted (to make space or something similar) and therefore not be restoreable from source. Over time the structure of those tables may change as well (with new fields only containing data for current records).
I would personally use Type I history in this scenario mostly (except for large transactional tables possibly). This will protect data from deletion and work in scenarios where you add fields. Incremental load is not that different from a technical perspective, but history is closer to the meaning of what you are doing. The added fields will not have content for records that no longer exist in source, but you can run a postscript or one-off script to fill in reasonable defaults if those apply. I would avoid the default column transformation in the short term as this does not deal well with changing data types (there is a bugfix coming for that).
As an alternative you can also load the old structure to a table and create a new table for the new structure with a table insert from the old table to load in the old data. Here you can also set defaults for new fields that did not exist prior. While this is very flexible, it is also slightly harder to understand.
If your scenario is different, then a different solution will perhaps be better
