EDIT: This post too was edited to be correct…
Forgive me, but I think you’re still thinking about it the wrong way.
If you load up a project and move your parameters around without writing any automation, you expect them to stay where they were when you last touched them, right? So if you move your rhythm guitar faders down to -6dB without writing it as automation, you expect them to be there until you move them the next time, yes? And this is regardless of if you close/re-open the project. Now, if you have VT turned on, without having written any automation, it should behave the same way. You touch and move a fader, and it should stay there until you move it again.
Now consider automation as just being the computer moving the fader for you, in a way that is predictable (because you told it to). So say you’ve decided to have the guitars at -3dB for the chorus, which happens a minute into the song. You’ve only moved the fader manually so far, so you go to the chorus and you engage automation and write -3dB for that section. Now you’ve moved the fader. Without VT the fader will return to -6dB before/after. With VT the fader will stay at -3dB after having passed the chorus. If you then touch it anywhere else it’ll stay where you leave it, it won’t move to either -3dB or -6dB, just wherever you left it.
So it should actually be entirely predictable. If you move the fader outside of where automation exists, then it stays where you left it. Once you read automation though - in the chorus that you automated - it’ll stay at the last value of that written automation (-6dB). Here’s an example:
VT: ON, automation -3dB at chorus.
Intro: You move the fader to -10dB and let go.
Verse: -10dB
Chorus: -3dB (automation read)
Verse: -3dB
Outro: You move the fader to -7dB where it remains until you press stop.
now you stop and press play again;
Intro: -3dB
Verse: -3dB
Chorus: -3dB (automation read)
Verse: -3dB
Outro: -3dB
When automation was written for the Chorus one automation point was written at the beginning of the timeline. That automation point is the same value as the first point written in the chorus - in this case -3dB. So when you started playback the second time it read that very first data point.
So as you can see the purpose of VT is not to make your levels 100% repeatable, for that you need to write automation either throughout all VT, or not use VT at all. Instead it’s so that you can avoid having to disable automation read every time you want to experiment with a section. So in the above example when you have already written the chorus automation, you end up with the same experience automating the verse. You can just move parameters around without them “snapping” to pre-recorded automation values. Then you punch that automation in for the verse and move on to the next section.
See how the thinking is slightly different?