I'm new to this board, but I work in this area, and I have recently come across three customers of the EDS Media Vault (based on Versant) who threw it out because it didnt work. Those three customers were AT&T GM a "large japanese entertainment conglomerate"
Part of the problem was undoubtedly the Media Vault, but there was negative feedback on the Versant database. Versant is great conceptually, but two main drawbacks are scaleability and performance in real world applications.
Who does Versant sell to, and where do they do well?
A friend of my just started work for Primus, supposedly a reference account for VSNT. They threw it out because it couldn't scale, was buggy as hell, didn't have any real client-side caching, didn't have a non-blocking read and wasn't any too fast on anything except batch loading of data. Before that he worked on the Sheriff project at MCI, also a VSNT account. They were having the same problems there too.
The Chicago Stock Exchange is using Versant - If it didnt work then one of the busiest exchanges should be told this. Sounds more like the application was written badly and they want a scapegoat. Versant is proven in too many companies for one experience with one customer to cast doubt on its viability. I have used it for a few years and its a top notch product that saves hours coding compared to Oracle, Informix, DB2 and the like.
It is hard to tell why the client had so many problems without knowing the application. As for MCI, they keep coming back to Versant. They are and have been a customer. As for scalability you should browse some of the news clips, like the one about the SABRE product by Air France. As for bugs try any one of Informix, Sybase, Poet, ... and tell me what choice the customer really has. Also, sometimes the problem is the customer. They take on huge projects without having a team that understands the issues. They pin themselves into an architectural corner, fail and blame it on the database. I don't know what the Media Vault project is, but if they were storing large binary objects in the database, without any kind of optimization (at an architecture level) they would fail. But this is easy to see for a DB person and hard to see for a Multi-Media person. Did they have their design checked?