Proactive Cache Connections

I have a running theory based on observation and I am looking for some opinions on the subject. Namely, that when a cube is configured to use Proactive Caching, when users establish a connection to the cube from a client app, be that Excel or SSRS, the client actually connects to the cache, not the cube directly.

Is that correct?

Here's why I am asking. I had a fact table that was set up to use proactive caching. The configuration used the setting "Update teh cache when the data changes". It also had the option to "Bring online immediately". With that setup, it also used a SQL Server Tracking Table for notifications.

Here's the observation...we had some SSIS packages that performed destructive loads of some of the fact tables that were configured as described above. The first thing the SSIS package did was truncated the fact table. Then it would run the source query and ultimately load th fact table.

While that process was running, users would complain that they lost visibility to the data in those particular measure groups while actively browsing the cube with Excel (or running a SSRS report).

My theory is that they clients connected to the cache. When the ETL ran, it truncated the table which triggered the cache to update...with no data in the fact table.

Am I correct in this logic?