Tier 1: Linux virtual consoles. Switch among these with Alt-F1 through Alt-F8. Control-Alt instead of Alt if in Wayland. I have seven with a text terminal and Wayland on the eighth. This tier supports showing only one virtual console at once.
Tier 2: Inside the Wayland virtual terminal, Sway managing virtual desktops. I use nonstandard keybindings here: Super-1 and -2 to cycle left and right, and Super-Q n to go to the n-th desktop. Beyond the first ten desktops, I can use Super-R to rename a desktop to a "named" desktop. For cycling purposes, these come after the first ten. This tier supports showing only one desktop at once.
Tier 3: Inside a Sway virtual desktop, windows managed by Sway. This tier supports splitting, showing multiple windows at once. I use nonstandard emacs-style keybindings, Super-F/B/N/P to move among those. These are often running a virtual terminal program, foot. I don't use a multiplexing terminal with multiple "tabs", because I favor a more minimalist setup with fewer tiers.
Tier 4: Inside a Sway-managed window, mosh. This tier isn't always present; I only use this tier if I'm using a remote system. Mosh has its own concept of sessions. These can be used in conjunction with Tmux's sessions
mosh's system is designed to smooth over connectivity issues. Lose network connectivity and mosh will display a message. Hibernate a laptop for a month with a mosh connection open to another machine, open the lid, and mosh will transparently re-establish its connections as if there had been no interruption. I mostly use mosh to reduce perceived latency, but the connectivity stuff is neat. Not much interaction with this tier, short of force-exiting with Control-^ . and this tier only supports showing one session in a terminal at once.
Tier 5: Inside a mosh session, tmux sessions. Tmux has its own set of sessions, which one can attach to with tmux attach. This tier only supports showing one session at once.
Tier 6: Inside a tmux session, tmux windows. I use a nonstandard prefix key for tmux (and GNU screen) to reduce friction with emacs
Control-O. I use emacs-style keybindings to cycle among windows
Control-O Control-N/Control-P. This tier does support splitting to show multiple tmux windows at once, though I don't use that functionality.
Tier 7: Inside a tmux window, I run a bash shell process. Bash supports job control. Control-Z to suspend the current job and return to bash, jobs to list jobs, fg %n to activate the nth job.
Tier 8: Inside a bash job, I might be running emacs, and that has emacs frames. If you're using graphical emacs, each frame corresponds to a window in your windowing environment. In terminal emacs, each is basically another invisible layer that you can switch among. C-x 5 2 to create a new frame, C-x 5 o to cycle, C-x 5 0 to destroy. This tier does not support showing multiple frames at once.
Tier 9: Emacs buffers. Each "buffer" might be a text file, a email client with mu4e, an LLM chat session with ellama, a "spreadsheet" with an org-mode table, whatever. One can show multiple emacs windows and assign a buffer to each emacs window (emacs has its own concept of windows, which kinda correspond to "panes" in most programs). Emacs has many systems for switching among these, but I mostly use one of two fairly vanilla add-on packages, either C-x b for ido-switch-buffer to switch among buffers using tab-completing names, or C-x C-b to use ibuffer, which provides menu-based selection.
Tier 10: Usually not something I use in conjunction with emacs, but if one is running a bash instance in an emacs shell-mode buffer (M-x shell), then bash's job control comes into the picture. Emacs shell-mode requires one to prefix each bash control key sequence with C-c, so C-c C-z to suspend the current job, and return to shell, jobs to list current jobs, and fg %n to activate the n-th job. Can only show one job at once.
EDIT: You could maybe make an argument that there's another tier between Tier 7 and Tier 8, because I use an emacs feature called desktop.el that persists an emacs session, including its frames and windows and open buffers and all across invocations of emacs for a given project. But I rarely use this, so it's not normally in the stack. If it's there, you can only have one active at once, no "split desktop.el" functionality.
EDIT2: I take it back. I had workspace renaming set up in i3, but never pulled that configuration over when I switched to sway. So just the basic 10 workspaces.