With DynamoDB you can batch process table updates and selectively send WebSocket updates through API Gateway, vs Firestore which sends everything and can only process records individually in order. Firestore pricing is cheaper (18 cents per GB vs 25 cents for DDB) but the bandwidth/retrieval costs are higher.

I used firestore in production and it scaled beautifully to 10k concurrent users (many physical classrooms with all students connected, mainly during US school hours). My main issues were storage/bandwidth costs which were substantial for a relatively high traffic application using Firepad (https://firepad.io/), and the cold starts/monitoring for Firebase functions. The trade-off for our small startup was clearly time - with Firebase we could turn a hackathon's worth of time into a fully-managed service that required virtually zero maintenance, and customize it to our specific problem quite easily (albeit exposing much of our application logic as client JS), while handling a handful of backend processes like payment processing with Firebase functions.

Nowadays you can build the same fully-managed no-maintenance service on AWS and get access to a far greater range of services and data centers. However they have by no means tackled the web console user experience like Firebase consistently has.