Metadata caching is used for queries which can be fulfilled directly from metadata, e.g. the row count of a table Query Result Caching is for queries that have been executed already and the results can be accessed by a queryID Warehouse caching is within the virtual warehouse instance and is usually based on queries that have already executed.

Correct. Temporary tables don't have any failsafe therefore there is no way to recover historical data

The underlying data must not change and the query should be syntactically identical in order for the query result cache to be used. The cache must have been generated (or last used) less then 24 hours ago. https://docs.snowflake.net/manuals/user-guide/querying-persisted-results.html#retrieval-optimization

Correct. The virtual warehouse will wait for the query to finish before entering the suspended state.

More then one worksheets can be opened on the Snowflake Web UI

If minimum and maximum value is set to same for a multi cluster virtual warehouse, Snowflake will start all clusters. The mode is called Maximized and is used to a) provide maximum performance b) statically control the resources https://docs.snowflake.net/manuals/user-guide/warehouses-multicluster.html#maximized-vs-auto-scale

External tables allow querying data in external storage without first loading it into a Snowflake table

Snowpipe uses a server less architecture and therefore doesn't use a virtual warehouse compute resources. Please see below link for further details https://docs.snowflake.net/manuals/user-guide/data-load-snowpipe-intro.html

Of these virtual warehouses and shares can't be cloned https://docs.snowflake.net/manuals/sql-reference/sql/create-clone.html

Incorrect. Failsafe can not be disabled for the Account or a database or a table. It is a Snowflake feature that is always present. The only tables that don't have failsafe (by default, and this can't be configured either) are transient and temporary tables.