The currently used TBX recreation fallback should be refactored in a way, so that the recreation of the TBX file from DB will be the default way.
That means for Import:
1. Import TBX internally, do not create a TBX with IDs
On opening a task:
1. Check the existence of the TBX file on disk, if it does not exist:
2. Create a TBX from DB
3. Set IDs in TBX either by doing it ourself or sending it to the termtagger (see TRANSLATE-376)