CQRS without event sourcing: a simpler approach?

we're exploring CQRS but without the full commitment to event sourcing, at least initially. the goal is to have separate read and write models to optimize for different access patterns and potentially scale them independently. our idea is to use PostgreSQL for the write model, and then stream changes via Debezium (CDC) to update a separate Elasticsearch instance for the read model. the read model would be optimized for complex searches and aggregations that are slow in PostgreSQL. is this considered a valid implementation of CQRS, even without an event store? what are the potential pitfalls we should be aware of, especially regarding maintaining consistency between the write DB and the read store? are there specific consistency guarantees we'd be sacrificing compared to a full event-sourced CQRS setup?