Actually, before jupyter I used a similar workflow (for python, but it doesn’t matter here). And when jupyter (especially jupyterlab) appeared, it was like fresh air after that.
I think there is some misunderstanding here - by “same layout between different computers” I meant the same files open in the same way, not just having the same editor settings.
This file contains inputs only, no outputs (at least for me). And of course there are no plots.
Sure, but you always need to manually choose which instance to use. While in jupyter there is a clear correspondence between notebook and backend instance, so they cannot be mixed up.
It sounds to me like the core issue is that with JupyterLab you are doing everything from a single server, and when you are using IDE’s you are doing it from multiple different servers. You can have the same reproducability with most IDE’s if you just did that from a single server as well.
Granted, if you’re connectivity is not good enough for an ssh session, that’s definitely an issue I don’t know how to overcome. If this really is a persistent issue for you you may have no choice but to stick with JupyterLab.
Is there a simple way to reproduce the exact window layout and list of open buffers between different machines? This seems to be also something OP wishes for.
I’m aware of save-desktop mode (or something like this) to reproduce the state when restarting emacs on the same machine, could one somehow share this between machines?
I haven’t tried it my self, but I think saving a perspective file on a remote machine should “just work”. There would probably be an issue if you access files by different paths on different machines, though (/ssh:user@myurl#20:/path/to/file.org vs smb:user%domain@myurli:/path/to/file.org or something like that).
EDIT: just tested this setup quickly, and it seems to work as expected.
