Ability to union structure-identical published data sources in Desktop, and the (optional) ability to publish this as a data source.
The inability of Tableau data extracts to update is a known problem, and this solution will create a simple, yet effective workaround for it, and I'll explain.
Most updates on fact tables are made only on a small fraction of the rows; in most cases, the recently inserted data.
TDEs do not provide a solution for this common scenario, except running a full refresh which might take a huge impact on server, networking and DB performance, depending of the magnitude of data extracted.
The proposed solution for this might look like this:
1. Create two identical data sources (even duplicate).
2. Change the first data source to extract, with a time frame filter to include all records which might be updated. This extract should be quite small, and should refresh quickly.
3. Publish this extract, set frequent refresh schedule.
4. Repeat steps 2-3, with the opposite filter. This extract should be quite large, and should refresh slowly. Set frequent incremental refresh, add a much-less-frequent full refresh.
5. Close the data sources (optional, but these DSs will be redundant in a second).
6. Create a new data source, connect to one of the recently published extracts, and then union to the other.
7. Ability to publish the data source, within the workbook or as a published data source.
It Is Needed
This new data source will be both recent and quick, and will have all data, something a lot of people were looking for:
and even in Stack Overflow: tableau - Merge/Union multiple data sources - Stack Overflow
A possible workaround might be to use the extract SDK (http://onlinehelp.tableau.com/current/api/sdk/en-us/help.htm#SDK/tableau_sdk_working_with_extracts.htm) but this has several downsides:
1. Requires a developer.
2. Needs to run on Tableau Server (which might be unreachable) / require a lot of networking if run on a different machine, as the whole extract should be published (uploaded).
Who Am I, Anyway?
I'm a veteran Tableau developer, administrator and accredited trainer, working since 2002 on various BI products and BI related products (DBs, ETLs, BI tools etc.).