This is something I've heard over the years but I'm not sure if it is still holds true. (I work with Oracle databases daily).
Oracle DBMSes are very robust, to be sure, and you can expect them to run forever. But in my mind other databases seems to have caught up and there really isn't a compelling reason to choose Oracle over other commercial databases anymore (also, Oracle pricing is a big deterrent).
In terms of performance, Oracle's licensing prohibits any benchmarking, but anecdotally I haven't found it to be particularly performant for most of my queries. Oracle used to be known for the innovative under-the-hood algorithmic improvements to the database, but lately I haven't seen anything too exciting, whereas SQL Server is getting better every year with new innovations (columnstore indices, in-memory features, Polybase, in-database Python/R computations, etc.).
> use cases I'd recommend it for over open source alternatives has shrunk to nearly nothing.
I agree, though I would say for heavy transactional use cases, I would still choose a commercial database over an open-source one. However, among the commercial databases out there, Oracle would be my last choice. It has too much legacy crud that have to worked around. The only reason Oracle is still around is because (1) In the enterprise, no one gets fired for buying Oracle. (2) fungible expertise in their services organization ensures business continuity, albeit at a lowest common denominator level.