An updated engine, damage displayed on sections of ship and to support organic looking things. More optimized from scratch.
I dont think the current engine can handle more detailed objects without crawling. Looking at objects in space stations look like 90 era
Regardless of how much the engine is upgraded, the modellers and artists can only work at the same rate. Elite Dangerous heavily re-uses assets because they only have so much time to draw them. A new game wouldn't fix that - and might well make it worse.
They could also upgrade the engine underneath the existing game, with a little bit of work on compatibility [1]. These things are supposed to be a bit modular, after all.
[1] They did just this for the sound components a couple of years ago.
Updated / reworked network engine. Seamless transitions. Instancec masked so that they aren't so noticable.
The only instance transitions that are noticeable are the ones in and out of supercruise, and that's likely as much an artifact of trying to connect up multiple clients to each other and have them all acknowledge, as anything else. And this stuff is getting faster - dropping out of orbital cruise is now a barely noticeable pause for me, whereas it was often a several second break when Horizons first released.
If you remember what Elite Dangerous networking was like in 1.0, a "start from scratch" reboot with a different instancing model is likely to have considerably worse performance than ED does now, for quite some time. There's no way to properly test performance with thousands of concurrent players with diverse network locations and hardware without actually having a released game running that network stack. (pre-1.0 networking was terrible, but worked absolutely fine in their internal tests...)
Total revamped gameplay story. A genuine storyline from start to defeat of the thargoid invasion. With many key characters and lots of lore / story elements.
As Frontier have stated, one of their big constraints on this is writer and translator time, which has virtually nothing to do with the game technology. They could write all the storyline in advance, of course - I hear players really like a railroad plot - but that would still mean a big delay before this "Elite Deadly" can be released - during which they either hire two separate teams, one to work on each - or just abandon Elite Dangerous development entirely apart from critical bugfixes for the several years it takes to get Elite Deadly ready.
I'd really like Frontier to double the size of their Elite Dangerous writing team - the game could easily find work for them - but it's unlikely to happen.
NEw engine to support atmos planets and legs.
Again, a lot of the bits are currently in Elite Dangerous - planets are in, atmospheric effects are in (stations, lagrange clouds, etc.) - and legs is just a matter of making a much smaller and slower SRV with a slightly different movement model. This is really not "new engine" sort of changes. The same "Cobra" engine underlying Elite Dangerous also runs their Jurassic World game, which is set on a planet with an atmosphere and things which walk around.
The work on actually having some
content on those atmospheric planets - design, models, textures, shaders, AI, documentation, interaction, gameplay, etc. - is going to take pretty much the same length of time (a very long one...) regardless of the underlying engine, but won't be significantly sped up by switching to a different one.
None of these are things which would go quicker with "Elite Deadly" than with continuing to develop the existing game. This is one of the reasons there isn't really much of a competitor to Elite Dangerous on its own grounds - by the time you've made a game which can do everything Elite Dangerous does, plus atmospheric landings and space legs, Elite Dangerous will have had several years to move on itself. It's not a great market for a competitor to enter, and it would be just silly for Frontier to try to compete
with itself on this.