The quick fix is to set the Trim duration (hard) start = 1083 on the strip for the 3D.
The root of the problem is that by default an action strip pulls its data from the start of the animation. You animated your text timed to line up with the start of the action strip, not the beginning of the scene.
The current Best Practice is to keep your VSE and your 3D separate. The VSE scene should only have action strips and no 3D objects. You can create a separate scene for the 3d animation that has no VSE strips in it. Then the VSE scene has strips that refer to the separate 3D scene.
If you're not in a hurry, you can duplicate the scene using the + button at the top and choose Link Objects. This action does not actually copy all the VSE strips, but it does copy all the 3D objects, so you can use it for the source of the action strip.