Hi Philipp
This is definitely feasible - we did something similar at my previous employer. The import package would however be set to import the same filename so that the parameters don't change.
Perhaps the process you use to deliver the file could also save another copy of the file with the date and time as a prefix to the filename? That way you have a history of data loads and you could manually reload an earlier data extract should something go wrong.
You could also create a bat file and schedule it to delete files older than a particular number of days in that folder.
For me the biggest issue is getting the actual data import correct so that it doesn't accidently clear other data e.g. journals, manual inputs, historical data. You can achieve this through different datasources, work status locks, etc.
Other things to consider would be which user ID to use to schedule the tasks and the use of a SQL table instead of a flatfile.
Regards
Dewald