The CS2 shuttle function itself has no extension for anything other than reverse (at each end) and an optional pause(s) in the middle

No , however you may use the same sensors for both the shuttle and a memory routine

The main consideration is that a memory is not specifically aware of which loco is in operation unless you tell it.
If you only run one train on the line then this is less of an issue , but still a sticking point.

Personally, I have given up on the Shuttle solutions and construct a group of memory routes which do the same thing but can add/coordinate additional functions (like sounds - ) and trackside animation (platform lights on/off, barrier arms down/up, car system stop) .

To address the possibility of using different locos, I would suggest designing your (proxy shuttle) memory routes to refer to an MU.
The MU would have a dummy head and tail loco in its consist , however you may alter a 3rd loco as needed into the MU rather than rewriting your memory routes

A screenshot please.

One aspect that comes to mind in certain cases/decoders is the need to turn off the function before it can be turned on again

Personally ...

I would only use timings in a reduced concise manner - never 40 second
e.g.
- telex uncoupling waltz
- coordinate startup from a platform - with Signal=Green, loco lights sound, platform announcement then Loco go.

For actions(sounds) en-route (like approaching a railway crossing) I would have another sensor and route.

I appreciate that examples exist (from Maerklin) where long term timings are used, however I consider these very fundamental and more a demonstration of capabilities than realistic or practical.

The very last thing you want is a 2nd route(memory/event) to start running when a previous one, involving the same components (loco, turnouts, signals) has not completed.