I was able to get the Cartesian orbital vectors for all the major bodies from HORIZON at the J2000 epoch only. I could extend the coverage forward thru time. It’s easy to get data overload doing this. My simulation is modeled using the Laws of Gravitation and Motion alone. This gives results that are surprisingly close to those published. Running the solar system backwards (by reversing the velocity vectors) has given me the initial vectors back to 1900. This is all I needed and the results were close enough for my purposes. I still have the CSV files.
I have also have had all sorts of problems with the horizons interface. For instance changing the date had no effect on the value of the vectors. i.e.: all specified start dates have the same values. Lately, I have not been able to duplicate this feat. There are obviously some serious problems with this interface, especially lately.
I know the data I got was correct because it correlates, perfectly, with published events, e.g.: the recent transit of Mercury.
I too am still looking for this type of data.