Elasticsearch and OpenSearch have a similar issue. They have a soft limit on the number of "searchable" fields (called "mapping") a table can have (by the way, a table is called an "index" in their terminology. How confusing!), which is 1000. We ran into this problem because we tried to jam all the logs in every single microservice into a single table. Each microservice had a different log format, so after combining them all the number of different fields surged. We did it this way because the operations team wanted to maintain a single table, by the name of simplicity, which to this day is a reason I can't completely fathom. We were surprised because we thought Elasticsearch and OpenSearch were some kind of magic box that can somehow ingest all the data we put yet still are perfectly performant. After that incident, we introduced a common log format that applies to every microservice.