We think option 4 looks the most appealing.
For database connections, large datasets, etc (stuff you can't send off to the browser) use redis.
This is a trade-off between DX and UX. Holding session state means you need sticky session routing, and restarting servers kills your user's sessions. Plus imposing a websocket on your users is a cardinal sin, makes scaling incredibly hard, makes page load times abysmal, makes disconnects a nightmare for both you and your user.
Its fine if you're creating something like stable diffusion web ui though, which is meant to be a single user app.
JSON is an alright storage format but it can’t do infinity, nan, or many other values important to Python. Also, you would have to write a reifier and serializer for every value. These simply create a problem rather than solving one.
A brief summary of what I did:
* Python-powered apps, like this
* React was used to handle lifecycle. Basically, I avoided writing any kind of lifecycle management because I just wanted to do exactly what React would do. This was achieved via events firing when React lifecycle events fired. If props changed on the server, I'd send the new props to React to see if it would trigger a lifecycle event. If a lifecycle event was triggered, I'd replicate it onto the server which could again render new props.
* The result was a nice seamless support for both native React components and the newfangled Python components (which could be made up of either more React components or Python components).
* The purpose of all of this was to cater to ML engineers who didn't particularly like JavaScript and just wanted to build their analyses with Python and share them with coworkers. We had an automatic deployment system for them to share these.
* Only caveats were around session management. Much of the state was handled on the server via a websocket, so if your connection dropped, you lost all ability to update the layout. Load balancing was going to be tricky, since the sessions would have to be shared somehow. I was thinking about serializing and deserializing using Pickle but never really got around to deployment.
See also: [Dash](https://github.com/plotly/dash)