The basic layout of the ship will remain the same. So most of the speculation below is off. However, some issues, such as where the solar arrays will be stowed, are still open for interpretation.
Material Change to Vehicle
The main change is that they are shifting from a very high tech composite to a "fairly heavy metal". The lightest metal routinely used in spacecraft engineering at scale is an aluminium/lithium alloy that was used for the Space Shuttle external tank. An unlikely option is something with a thin layer of stainless steel, which would be necessary for a reusable vehicle using liquid hydrogen. That said, bear in mind that the initial flights to Mars may need liquid hydrogen in reserve (rather than drilling for water) because the reserves of water would have yet to be proven on Mars. This was actually a key design detail of Mars Direct, which was designed before glaciers were observed below the dust in middle latitudes. These glaciers are ideal for providing propellant, but it would take a crew to set up a proper drilling rig and return the melted water to the surface for processing. The first ships will not have crews, meaning that either they have to take a fairly hard risk and build the propellant plant on the first crewed mission or die on Mars, or they have to have a backup plan like a hydrogen tank. A lot has to go right on the first two missions to get that infrastructure in place. That said, there would be no shortage of volunteers.
They also announced that we will be seeing "cool pictures of the demo Starship that will fly suborbital hops in the coming (roughly) 4 weeks." Four weeks out from the announcement date would be January 5, 2019.
SpaceX applied for a license to test the "up and down hopper" version of Starship in Texas. They will initially do short tests up to 500 meters/100 seconds. They will then extend to 5 km flights up to 6 minutes. They are also finalizing construction of the test facility/launch facility with a large tent, fuel tanks, antennas, and moving a lot of dirt around to make causeways across the soft sandy ground.
(Original Article below).
Recently, Elon Musk mentioned that the new spacecraft in development, known originally as Mars Colonial Transport, then Interplanetary Transport System, then BFS, is now being called simply "Starship". The booster is now also simplified from BFR to "Super Heavy". This makes some sense in that his satellite constellation is being called Starlink. This seems to go back to calling the space-suited mannequin on the Falcon Heavy test launch "Starman" after the David Bowie song. It wouldn't be the first time - Dragon is named after Puff the Magic Dragon, and Falcon after the Millennium Falcon from Star Wars. The two drone ships are named after sentient starships from a science fiction book series. There are also lots of Hitchhikers Guide to the Galaxy references in his nomenclature.
He has also indicated that he recently-announced Hello Moon version of the Starship design is being changed yet again. The new change would be counter-intuitive, and when asked, he simply said, "RADICAL CHANGE". He also said the new design is exciting and delightfully counter-intuitive. Note that he also said the Hello Moon design was counter intuitive due to the front and rear fins rotating in terms of lateral drag but appearing superficially like forward control surfaces.
Keep in mind that Blue Origin has second mover advantage here. Whatever SpaceX settles on for a design, Blue Origin can come up with something similar or better with New Armstrong. Constantly changing the design in radical ways will tend to kill that second mover advantage for Blue Origin in terms of years of lead time. If SpaceX had stuck with their original design, Blue Origin would be studying it extensively and have a three year lead time in coming up with something better. As of now, they are just as lost as we are. That may also be part of the purpose of the ever-changing design, and some of the Issues mentioned below.
Analysis of Issues
Here are the key issues I've noticed with Starship's design...
No escape system - It is designed like an airliner, but with no escape system for ground evacuation or launch escape.
Solar Panels - Most recent versions do not seem to have a place to stow solar panels. The ones protruding from the initial design could have been stowed in the tri-form fin-like elements, but the past few ships show nothing but tanks in the rear sections. This implies some sort of lateral spine or (in the Hello Moon version) a place to put much smaller panels in the tail-fin.
Center of Gravity/Center of Pressure on Landing - The Falcon boosters can land tail-first in part because all the engine weight is in the tail. They basically land like lawn darts. A Starship will launch with a heavy nose, fuel tanks, and a heavy tail with the engines. After using up the fuel supply from launch, there is still weight in the nose and tail, averaging out the center of gravity. That said, the rear fins went from protrusions to a small delta control surface to a full pair of fins because the side-facing atmospheric entry would make it difficult with so much mass in the tail to keep the vehicle stable.
Unloading on the Moon/Mars - Depictions show a door opening out of the cargo bay side with a crane lowering components to the surface. This would work with some elements, but may be difficult with large equipment. It would also be very difficult with passengers in mass or any sort of emergency landing with a point-to-point flight on Earth. Unloading like this with heavy gear may actually risk tipping the ship over, especially if it were landed on any slope toward the cargo door or an unsteady surface.
Engine Plume Hammering on Unprepared Surfaces - One key reason that NASA resorted to the extreme measure of landing the Curiosity rover using a skycrane was that engines big enough to land a one ton rover could easily drill the lander into the surface by way of the extreme supersonic exhaust plumes of the rocket engines being so close to the surface. The Apollo LEM dealt with this by using a wide engine bell, low impulse propellants, and cutting the engine about a meter in the air and then dropping down to the surface. Existing Mars landers typically use a cluster of tiny rockets in pods on the side to spread out the exhaust plume, and also use low impulse propellants. SpaceX is using fairly high performance engines, and sea level ones in the case of the Hello Moon version, for landings. On Earth, landing on a concrete pad solves the problem for the most part - the surface is strong enough to support the lander on a level surface, and also to take the force of engines blasting it with hypersonic, superheated propellant. Unprepared surfaces would have to be very level bedrock outcroppings.
That Massive Window - This would be heavy, fragile, expensive, and deeply impractical from a structural prospective. The only aircraft I've ever seen with a cockpit like this is the US B-36 bomber. While the second version of Starship lacked this window, the Hello Moon version brought it back again.
So what does that mean? SpaceX must be aware of all the issues listed above. Here are some speculations off the top of my head - some wild, some pragmatic. The wilder ones are simply in response to how loaded that announcement was in terms of language.
Canard Changes? Given the purpose of the canard wings on the Starship, it seems logical to change the arrangement from rear-swept to forward-swept to decrease drag and increase "reach" for the movement arm of the control surface. If the lateral engine nacelles were used as noted above, a second set of downward-angled control surfaces could be on the back side.
Move the Solar Panels up to the Cargo Bay? This has a lot of potential. They wouldn't be subject to the stress of being so close to the engines on take-off. They would also be more accessible by the crew in the event of a deployment or stowing error. On the surface of the moon or Mars, they would be up and away from the dust and surface operations close to the vehicle. This wouldn't be a radical change, but it's got some practical benefits. They could also be more practically scaled to the mission by being larger for deep space, or possibly replaced entirely with a low-enriched uranium reactor like KiloPower.
Detachable Payload/Crew Section? This also has potential for Lunar and Mars operations. One benefit of Mars Direct is that the Earth Return Vehicle (filled with propellant) is stored a safe distance from the crew habitat. By recombining the two (the very first concept that lead to Mars Direct was a single vehicle like Starship), SpaceX adds this risk.
Also, habitats can be left behind or used as space stations, particularly if they have their own solar power supply as mentioned above. Note that the current Starship design has more cubic meters of pressurized volume than the entire International Space Station. It simply lacks docking ports and would need a much larger solar array. A detachable payload section would have both. At any rate, ISS replacement could simply be a crew module with a solar array, coupled to a power/propulsion/docking Service Module on a second flight. The Service Module would also have a lot of potential working in series for space factories, solar propulsion systems, and fuel depots. It would basically eat the entire Blue Origin and ULA long term road map in one swift stroke.
This would also add some capacity for crew escape, even if it's a passive detachment with parachutes or landing rockets.
Lateral Engines? This is pretty interesting. The Skylon design has engines on the side because it was derived from the HOTOL concept from a couple decades earlier. HOTOL would have been a European space shuttle, and used a very heavy engine arrangement at the rear. However, the center of gravity would have shifted too far back to be stable after fuel burn. This is also a problem with Starship, and had lead to the ever-increasing wing size at the tail. Having the small landing engines on nacelles off the side of the vehicle would deal with landing on unprepared surfaces. It would also give an option for crew escape if coupled with the smaller propellant tanks used for landing in the original design. It may allow for cargo at the rear end if all the engines were put on pylons. It would remove some of the need for the rear wing, which may be difficult to engineer due to the stress on the hinge and electric motors. It would give a deployment platform for a crane and a counter-weight for laterally unloading payloads. It may have also led to the name change. After all, nacelles are a design hallmark of the Enterprise and Firefly's Serenity. Elon also doesn't design anything unless it looks better than Hollywood could design. This could be a pretty elegant design when he gets through with it. With such inspirations as Serenity and Enterprise, not to mention the Star Wars Y-Wing, that could be the inspiration for the name change as well. It certainly would look like a starship in the popular imagination. The disadvantage would be drag on takeoff from Earth, but it would allow for retractable landing gear that could be scaled to the engine arrangement and destination. I can call this neither probable or improbable, but simply file it under, "I wouldn't put it past them".
Payload section at the back? While this would solve the load/unload issue, it would make crew escape much worse. It would also make center of gravity far worse, as well as subjecting the crew/payload to intense vibration. This is not entirely unlike the back of the Shuttle cargo bay being close to the engines, but it seems incredibly unlikely. It would also be relatively ugly on the interior, since putting windows that close to the compression between the tanks and the engines would be suicidal. For these reasons, I'd say no. Though a variant of it is already in place with the tail-end cargo pods for the Hello Moon version. I cannot see them doing more than that. Then again, if they put all the engines in nacelles and not just the landing ones, this may make more sense. The cabin could simply drop down to the surface.
Split Payload Transformer? Imagine a ship with a payload nose section that splits in two. Each section could be lowered to the ground for easy access. The side with the crew/passengers would also have an escape system. It could also be split into more than two sections, provided they had some symmetry. They could even be spun for low-level artificial gravity. This seems highly unlikely, because having multiple pressurized bays becomes incredibly heavy for no practical benefit. I'd say unlikely for this reason. Elon Musk loves platonic ideals, not complexity for its own sake.
This rather extensive and somewhat wild flight of imagination is the sort of thing one pictures when Elon Musk says, "radical change" and "delightfully counter-intuitive". I suspect it will solve some of the issues noted above, because they would need some dramatic reason to change the design at this stage. Elon Musk likes idealized solutions with minimal draw-backs, and tends to iterate towards a democratized, ubiquitous product line rather than away from it. I may have hit on some of it with the more likely options.
Knowing SpaceX, I'm looking forward to being proven wrong in some delightfully clever way.