Creating databases with shared throughput is not reliable

I create 4 databases (under the same Cosmos DB account) back to back programmatically using your .NET SDK for Cosmos v3. It seems there is some kind of bug with shared throughput per database.

const int throughput = 800;
await cosmosClient.CreateDatabaseIfNotExistsAsync(databaseId, throughput);

The SDK sometimes creates the first database with 800 shared throughput, but it will ignore this request for subsequent databases. So the other 3 databases will be created with dedicated throughput per container. Now, what’s strange is sometimes the SDK works as expected and creates 4 databases with 800 shared throughput each. I just got slammed with an unexpected bill because of this. I was also able to reproduce this several times.

[bartelink]

Are you perhaps exceeding the max 25 containers with shared database throughput limit? (IIRC you can still add containers at that point, but only with RUs allocated at the Container level - perhaps your code is specifying the limits at Database and Container level?)

[msftbot[bot]]

Closing due to in-activity, pease feel free to re-open.