The type of GUID I am referring to is the COMB-style so that it can be used as part of a clustered index with minimal performance impact on inserts. More detail about COMB style uniqueidentifiers can be found here:

Articles

I am interviewing for a job and apparently they are having a ridiculous amount of conflicts being reported in Merge Replication. Unfortunately I do not have all of the details of their environment. The PK's (integer) are throwing the conflicts.

Assume millions of records in a table, should I be worried about page splits?

Assume billions of records in a table, should I be worried about page splits?

Any other concerns?

Also I will try to research this on my own, but if anyone knows: does table partitioning work well in a merge replication environment? I don't think the client has currently implemented table partitioning.