My strategy is using a separate, but equivalent schema for every tenant and setting the tenant context with every request before running queries.
Customer A is running the new schema, so gets one cluster. Customer Z will get migrated in the next hour, and then none of the old system is running.