Describe the problem

In #98385 we saw the default span config being used in the replicate queue for a short period. During this period, replicas were removed from ranges due to the allocator believing the replication factor should be x3 instead of the real factor x9.

This points to a wider issue where it is possible to use the default span config in place of the real span config. This condition is checked for in the merge queue , however not in other places that should:

• replicate queue
• store rebalancer
• split queue
• mvcc gc queue

To Reproduce

See this comment.

Expected behavior

The default span config is never used in the above places so that incorrect decisions don't occur.

Jira issue: CRDB-25257