How to use Incremental Refresh in Dataflows?

Hi community,

More silly questions still trying to understand fundamentals here, but just not quite getting it.

I have a Workspace for 'Ingestion' entities, pulling data from an on-premise datasource (SQL Server).

This ingestion Dataflow is running on Incremental Refresh - it took around 3 hours to do the initial load, but now consistently refreshes on schedule in around 4 - 6 minutes. Great so far.

I then have a 2nd Workspace for staging ie. linked entities, with Calculated entities referencing them. (I'm still unsure if its better to do this all in the 'Ingestion' workspace, or to do staging / ETL in a seperate workspace like this.. that's another issue but any suggestions welcomed 😀 😞

The 'Staged' Dataflow is however taking excessively longer to refresh than the 'Ingestion' dataflow, around 1 hour an 45 minutes:

So I'm thinking - 'this time should be at least 'comparable' to the previous  incremental refresh time.. Do I need to set up Incremental Refresh on my Calculated Entities in this Dataflow?'

However, this Microsoft article seems to indicated that Computed entities behave the same way as Linked entities, which don't require incremental refresh (as Linked entities are simply a pointer), and that Calculated entities are simply performing queries over the existing stored data, not 'storing' the data again within themselves..  so putting incremental refresh on Calculated entities  doesn't seem to be the correct method:

So.. my 3 big questions are:

1. Does anyone know how this should be correctly configured and why my 'staging' refresh would be taking so much longer than the 'ingestion' incremental refresh?
2. Do we only use Incremental Refresh on the Initial set of Ingestion datasource entities, or do you need to apply it on every 'step' of the way when working across multiple dataflows and/or multiple  workspaces??
3. If linked entities (in a separate workspace) are effectively are already pointing to previous dataflows/workspaces, and calculated entities simply reference those linked entities.. then why do we need to refresh these subsequent dataflows at all?  (I am aware from this Microsoft article that dataflows in previous workspaces are treated as an 'external datasource' and therefore apparently need refreshing.. but how does this work when it appears that computed entities are in reality just referencing linked entities (which in turn are referencing the 'source' entities)... i visialize this like they are still just all 'pointers' in a series.. so why is a refresh even required on calculated entities?)

Thanks for sticking with me... any help greatly appreciated!