I want to see this running on Xbox One.

What I want to see is it in action, in a game, that then, while running, the game loses its connection to the cloud and continues to run.

I am very interested in seeing the transition between those two stages (to and fro)

Nah dont see it that way.

Making games and testing games is very hard work. To actually have to create two states, one online and one offline is too much for most if not all developers to contemplate. Can you imagine the work/decisions you would have to do to decide to use the cloud and then have a fallback situation if the player cannot get online... would the game then look/play worse in offline mode... how do you communicate such a thing to the player.

This is useful for online games no doubt, but i expect no 3rd parties to use it for anything unless it's a online/multiplayer component. It's too much effort and too time consuming. In a time when games are getting more expensive, it's just not going to be a benefit.

I have tried to explain this many times but people are so locked on terms they miss the tech behind the scenes.  Calling Azure just dedicated servers is simplfying what actually goes on behind the scene.  People dismissing the term cloud do not really understand what type of cloud platform MS is using.  For people that are interested just research PaaS (platform as a service).  This is what MS is offering and dedicated servers are part of the cloud platform.  

As you mention in your post, there is no server up and running waiting for clients.  When the first request comes from a client, Azure spins up a virtual server.  That server will remain up for as long as there are clients hitting it and will drop when no more clients are hitting it.

Here is the thing about such a system, its only expensive once.  Once the engine is created to switch between the 2, you probably never have to really code for it again.  The reason why is that you have your pipline structure to either retrieve data from the cloud or locally.  Its one of those type of process that engine developers work on between projects.  

So I agree outside of developers that are producing 1st party or 2nd party games, we probably not see many games using the cloud but later in the lifecycle of the X1 you probably will see a lot of games go down this route as engine developers will have time to incorporate the benefits within their engines.

Well that's not entirely true :) yes the bulk of the code work would be done for whatever you are writing (communicating with servers, authentication), but this still has to fit into the game, not take up too many ms, and work with other systems. Each game will have their own strengths and weaknesses. Actually it's not that difficult to get something working with an online server, its making it work with all other systems within your game and making it improve your game - thats where the cost/time and issues really create problems.

If you were making a game say it was Forza, and you wrote the game to do weather simulation online, if you made Forza 2, it is true that would be slightly easier as its probably using a similar structure and engine as before. If you decided to make a totally different game, all bets are off. It's like starting again.. and theres no guarentee that you would still have a benefit from it once you had got into development of that other game.

It's really difficult juggling all these things... developers dont want the hassle. They have many other priorities and I dont think this will be one of them. I dont see many 1st party using it in the way it's being sold. For example Sunset Overdrive is using it for Matchmaking, which is sensible, but that's not making the actual single player element any better. So as i said, you will see it for MP/MMO benefits and not really in SP games. That's not to say its not a benefit, but not in the way that it's currently being sold.

One thing this developer states is how they are using the cloud.  If you notice none of it really rely on ms it takes to talk to the server.  All the things listed are global effects which can be process async and not dependent on what the client is current is doing.  I agree if the developers are going to use the cloud for anything reliant on the client performing an action and then processed by the cloud would mean that your connect to the server must have a low ping, low packets (sending binary helps).

On your second paragraph, most of your code you would reuse.  Here is the thing, if you are processing some type of physics operation, the code does not change when run on the server compared locally especially if the CPUs are x86.  Its the same code run server side as its running client side, the difference is your pipeline.  The code that switches between one or the other dependent on whatever triggers used for the switch.  That’s the part that’s a pain in the ass.  Once you write that code then you just write the physics part once and reuse it either server or client side.

One way to help with juggling this type of code is the Orleans platform MS created just for this type of scenarios.  MS actually has thought about the complexity of doing cloud based processing and spent 3 years getting that Orleans platform up to make this process much easier.

Making a game is a very complicated process, there are many different features fighting for ms and performance - if a company is making a game, mostly they go for the path of least resistance and the easiest option. Not only is it cheaper, its faster. If a data call suddenly fails, the game would have to react to that very quickly, so you cannot really use this system on something that needs fast response.

It may not be difficult to call the servers, send and receive data, but getting this whole process that it works in a game and then also figuring out how to make it work when the system is offline, its just too much bother for people making single player games. That's the problem. This feature as I said before, it's main benefit is online games, matchmaking etc. I do not expect the majority of games (i.e 99%) to use it in SP games. If that's the case then it's of no benefit to most core content of games.

But we shall see, the proof is in the pudding.

The proof is already out there.  Distributed or as some know it client server technology has been around longer than video games.  There is nothing new here and its not like this type of stuff does not happen every day.  

I do agree that 3rd party developers will always go for least resistance but I am not talking about the now.  Big time developers will improve their engines over time because engine developers do not work on the current engine during game development.  You can also believe that MS will be sending their people to help with integrating cloud tech into current teams engine just like they sent people to Bungie and other developers to help with working with the ESRAM.

You are saying its difficult and I am telling you its not difficult its time consuming.  You keep saying this is only good for online or matchmaking etc but you already have proof from MS own studios as well as others.  Anything new takes time to work into existing projects so yes, we will only see limited uptake in the next few years.  Once the technology is shone to be sound, you can believe more will use it as there will be a lot of avaliable code and MS own library avaliable to push the tech.

In the end I guess we will see how this goes, but in about 3 years who knows if we can have this same discussion.