Eclipse must be the buggiest software I've used in years. When you start working with large and complex projects, things just start to fall apart. It keeps running out of memory, projects randomly fail to compile, forcing you to restart Eclipse. Sometimes your project compiles with no errors, you then make a change that causes complier errors, you undo the change, and yet the project no longer wants to compile, no matter what you do. Once again, you need to restart Eclipse.
Or the project fails to compile, you go to the location where it claims the error is (this in a file that you haven't even checked out yet), but as far as you can see, there is no error. So what do you have to do? Make a dummy edit somewhere in the general vicinity of this phantom "error", undo the change and then recompile. Now it suddenly starts working again.
And don't get me started on all the bizarre error messages. Like the classic we get from time to time: "An internal error occurred while trying to display an error" (paraphrasing).
And yes, the ecosystem is pretty extensive and useful when it works. Example:
We needed the ability to consume a web service in Java (create stubs from WSDL at compile time given a web service URL), and for a test service I used ServiceHost .Net 4.0. It took me literally 20 minutes tops to create the service in .Net that exposed a simple Hello World method.
Then I searched around for a way to consume the web service in Java, and initially settled on Axis2. After a huge amount of effort and hair-pulling, we discovered that the reason things just didn't work was because there was a bug in the Axis2 OSGi bundle and some dependencies were not properly included (Eclipse tries hard to make it easy to manage dependencies and versioning but when things go wrong...). The whole point of an OSGi bundle is to be a self-contained bundle that can resolve all dependencies.
Eventually we gave up on Axis2 and went to CXF. After we got that figured out and the project in Eclipse configured to automatically pull in the WSDL and create the stubs at compile time, things started to work better. However the amount of code that was automatically generated from the WSDL was mind-boggling. Not that it really matters but compare to what you get with WCF. Also using the service in Java is not straight-forward, as the structures you exposed in the service are wrapped in special classes.
So yes, the "ecosystem" is pretty extensive, but in many cases those bases are already covered pretty well (e.g, WCF) when using VS and you don't need to relay on flaky external libraries.
EDIT: I didn't watch the presentation because I find attempts to move everything into the browser to be boring, pointless and a huge step backwards. But I have to wonder, was it explained how moving something as complex as Eclipse "into the cloud" is going work any better when Eclipse can barely function as a native application right now? Was it explained? And what about things like multiple monitor support, etc?