Quite a number of things but I’m scared to list them for fear of someone here thinking it might be a good idea to play with it.
Things that oracle does right: high availability, consistency of data (which is harder than you think, and the tools we use tend to use are genuinely crap at this especially things like MySQL and MongoDB which are insanely popular), performance of OTLP workloads, and support*
But Postgres and MSSQL are the true equivalents, you’re right.
Correctness is hard was my main comment. There’s a whole class of other database types, but oracle can do NOSQL document storage just fine.
Postgres and MSSQL can too, especially the former, but there are some weird quirks, like having documents larger than 16KiB means they’re stored off-row in a compressed TOAST table. Which hurts performance a lot.
Mysql, Postgres and Oracle are closer than ever in functionality, but I don’t think postgres/mysql alone can compare to “Oracle”, but if you compare postgres+greenplum/citus/XL or mysql+Vittess to Oracle+RAC that’s pretty close, but you’ll probably need to choose from a few extras to get all the way there. The key difference is Oracle is doing it all in one monolithic stack.
Oracle isn't really monolithic though. If you dig into it, it is all a really ancient aggregation of various tools and components from different eras. All of it has some Oracle branding on it, but usage-wise and functionality-wise it is inconsistent as hell, overlapping, downright weird and sometimes even dangerous to the non-wizard. Imagine the ancient font-installation dialogue in windows, times a thousand.
For most (but not all) purposes, PostgreSQL is approximately as good, the devil is in the details of your specific application. Oracle is a competent OLTP system but it makes a different set of tradeoffs than PostgreSQL. For geospatial performance, as an example, neither platform clearly wins -- it goes back and forth depending on the details of what you are doing.
The tl;dr: start with PostgreSQL and evaluate from there.
Things that oracle does right: high availability, consistency of data (which is harder than you think, and the tools we use tend to use are genuinely crap at this especially things like MySQL and MongoDB which are insanely popular), performance of OTLP workloads, and support*
Your mileage may vary on that last point.