I think that this is paradox because MIDI is not audio. What you want is to treat MIDI like audio. A freeze is a audio bounce in the background. You then want to manipulate this audio. But you can already do that. You can just make a mixdown and then manipulate the audio.
Why is it not sufficient for you to make an audio mixdown and then to manipulate the audio?
Why exactly do you want to have this feature? What is the problem that you face?
Let me outline what I think is paradox about manipulating frozen tracks. Reproduce this in your imagination.
Lets say you have a short 1/16 C1 note of MIDI at the beginning of measure 1. This MIDI Data triggers a long sample, lets say a long two bars GrooveAgentOne “splash” sample.
Now we freeze this MIDI track with let’s say a 8 seconds tail.
So now we can cut between the first and the second bar our frozen midi. We drag the second bar of the splash to lets say bar 8.
Now we have the beginning of the splash from bar 1 to 2 and the other half ending at bar 8 to 9.
Ok, now unfreeze the track…
…
…
…
How do the acoustical events , the split splash, translate back to MIDI in conjunction with our drum sampler?
Maybe it would make sense to outline your usecase and workflow so this community can come up with a solution or think of a possible feature that could assist you in such a situation. 
Cubase Rocks!
Gr,
JHP