We should provide the correct errors and prevent replication from starting when a user tries to:
replicate mismatched table names.
a user tries to replicate a system table
upgrade a cluster during an initial scan
run sql on the cluster during the initial scan (maybe this will be irrelevant because it'll be online)
start replication on column family setups or other table configurations that we don't support. when I did this during acceptance testing it just forced everything to go to the DLQ and that was not great. could we just error this out instead?
We should provide the correct errors and prevent replication from starting when a user tries to:
Jira issue: CRDB-41356
Epic CRDB-41465