Now, that's interesting! I'd be really curious to see how you do that.
Simple, I am connecting to the source sever, whatever platform it may be, querying it, or downloading if is an FTP or a website, and then loading/inserting data to the target server, usually SQL, and than do whatever I need with it. Sometimes I do transformations inside Python if they can't be done in SQL, i.e. SRID projection conversion for spatial data. I also handle the errors and the logins the way I want. I simply prefer to write simple code instead of dragging and dropping and then "painting" arrows... I built some simple modules that I reuse when I connect, executing queries, manipulating files, which have logging incorporated and emails sending if somethings goes wrong. SSIS is a fine platform but it is kind of complicated especially when it comes to debugging, deploying.
In regards to IronPython, it is too bad that they haven't went to use Python 3 yet any only for SMO I would not switch to it even though it would be very nice to have access to. Maybe somebody will build some SMO module in Python 3 and you wouldn't need IronPython anymore....
I am not an advanced Python developer at that level just a DBA. No need for irony here.
Don't just give the hungry man a fish, teach him how to catch it as well.