suggestion

Is it possible to extend the message table with a field node_id to be able to group duplicate messages from watches/replies together?

A poor man's (and maybe faster to implement) approach is to have a LIKE statement which is searching msgtext for the %[id://NODE_ID]% text.

Actually I'm not sure if grouping should be done on the writing (avoid duplicates) or reading side.

Having to delete multiple duplicate messages is a bit cumbersome, and especially fiddly on mobile.

That's easier solved by avoiding duplicates.

Update

Having a field node_id would also allow listing all messages concerning a node together with the node's display.

Update

A third approach: I'm thinking about writing a display filter which is parsing and grouping the messages. This could also be done as Nodelet hack 🤔