Just ran a basic test of this shipping idea for the map (960x640) I was thinking of. Set stuff up, and just let it run through the 25 years on max speed while I did something else.
Since I don't have the actual ship models exported yet I used one of the game's default electric locos (2D2) since it was about right for speed. That worked ok. It looks like even with fairly low speeds by land standards, it's still possible to make long hauls across oceans reasonably profitable. The longest run was from Southampton down to St Malo. Even that was making decent money, despite St Malo being a bit out in the boondocks and the trip taking several months of RT time. This is good, and means it's not necessary to go with insane speeds for ships on largish maps.
I also tested using a setup event to set oil and sand consumption to -100%. The idea here is that the ships will be slow enough anyway without having to stop for stuff all the time, and it doesn't make sense for them to need sand anyway, so just keep their stops to basic station turnaround time and forget about topping up the oil. It will also save clutter around the ports/stations, which are going to end up quite cluttered already. IRL ships do thousands of miles without stopping for oil anyway, so I figure they can do the same in RT3.
What I found was that setting oil consumption to -100% doesn't actually reduce it to zero. What it does is reduce the rate of consumption to between 1 and 2% per year. I think this is another coding safeguard against divide by zero errors, similar to the way the possible -100% cost on Connies in the Chile scenario only ended up bringing them down to $2k instead of zero. That's still fine, since over a standard scenario length it will only chew through about 40% of the initial oil, leaving about 60% for good reliability over 25 years or so.*
I'll get a fairly basic playable hull model sorted soon, so I can double check the hiding of the consist inside the hull around corners, and also see if there are any other problems I haven't thought of. Given the size of these things they are going to be beasts to skin within the constraints of a 1024x1024 image, but I think it's doable.
Short version: this ship concept is going to be workable in play, once a few minor skinning issues are sorted.
PS: Had a brainwave. There's no need to do ships as electric locos. They can be done as steam or diesel. In fact, for pre-1904 shipping (which someone will want to try sooner or later) any ships would have to be done as steam or diesel anyway since the game won't recognise the existence of electric track before 1904. It allows you to lay electric track via the editor in 1830, but for loco-running purposes treats it as normal track until 1904.
So, the new plan is to just use "electric track" to lay down the shipping lanes, so that it can be skinned as ocean rather than as railway track without affecting the normal land tracks, but to not bother about electric locos to run on the sea tracks. The ships can just be steamers, as they would have been, with all necessary adjustments and with locomotive availability set by territory. This will be the only way of doing it pre-1904, so might as well do it that way for any period.
*Oh yeah, some other things I found when checking trains at game end, as well as keeping an eye on breakdowns during the game. It seems that faster locomotives break down more often for the same nominal reliability rating, and also that faster locomotives and/or heavier locomotives chew through oil, water and sand faster than slower and lighter locomotives. The difference isn't enough that you would usually notice with standard length runs between stops, but when testing over decades with no maintenance stops it becomes obvious.
For example, with the "ship maintenance" event just being a blanket case for this first test, and applying to all locos everywhere, a P8 will run for well over 30 years while pulling full 8 car consists, without running out of water, while a Schools class will be out of water after about 27 years and will break down far more often than the P8, even though both have the same reliability rating. The reliability rating of the Schools class also deteriorates faster, reaching a level of breakdown probability after 27 years that takes the P8 takes about 33 years. Berkshires are even worse on water, sand and oil, and are also a lot heavier, but are about as reliable as the Schools class, which I put down to the Berkshire being a lot slower.
This would also explain why the default Mallard is such an unreliable piece of crap even though its nominal reliability rating isn't too bad. The thing goes like a rocket, so if the game is coded to penalise fast locos it makes sense it'd be unreliable.
Got it sorted. Turns out the game is fine with the "loco body" being shoved way back relative to the 0,0,0 point. This fixes the consist attachment point. 
