data lake store data predominantly in parquet format , whereas profile store data in columnar database (could be cosmodb). also identity graph is stored in graph database. this could be the reason for higher storage size in profile store compared with data lake.

if you are questioning the guardrails, best to try and break the guardrails and see what happens 

if you consider independent ECID attribute as identity, you will end up marking this  attribute as primary identity, because there will always be anonymous browsing events which will only have ECID as identity. which also means that when it is login browsing still ECID will be primary identity. As a...

if you trigger your journey based on audience qualification, then the profile must have stitched already. 

from your question am assuming you are ingesting into individual profile class dataset. from RTCDP perspective, there may not be much impact. profile store will refer to the latest record of the profile as ingested.  from Data lake perspective, yes, there will be impact that huge number of records w...

 using marketo engage source connector, creating a dataflow, there is a toggle button "sample dataflow". As per the documentation it says that this toggle button helps to restrict historical data and can be used for testing purposes.  But is there any impact on handling incremental data with this to...

 is there a way to disable data flow partially, meaning that fine to process the data to data lake, but not to profile store.  source connector is httpapi, target dataset is profile enabled.

As far I know this is not possible. You can append timestamp/hardcodedtext/audience name and few more standard text.if you want each audience to go to different folder, then you need to create as much destination.