I have a distributed application consisting of many components that communicate over TCP (for examle JMS) and HTTP. All components run on internal hardware, with internal IP add
There is no need for you to use an external public CA for a closed community project. In many larger organisations they operate an internal PKI to issue certs for internal projects like this. An advantage of using a PKI is that you can setup a trust relationship between the various components based on a single securely distributed root certificate / trust anchor.
However, if the project allowed internal users to connect securely to an internal service via their web browser you may want to consider using a public CA issued cert. The alternative is to make sure that every browser that may need to connect to your service trusted your root cert; this is to prevent browser warning messages.
I'd say it's reasonably safe, unless you think a ninja infiltrator is going to swap your server on you.
The 3rd party is there to make it harder to just 'up & generate' a new cert. Someone could re-create a self-signed cert on a new machine with the same details, it wouldn't be the same cert, you'd have to add an exception for it too, but your users probably wouldn't know the difference.
As long as your system is running inside your group and there are no plans to expand it (and plans do change, so keep that in mind), it is just fine to setup your own simple PKI infrastructure.
If you do end up expanding beyond your organization, all you need to do is distribute your root certificate to the parties you will be communicating. This gives actually a fine grained control to your partners how much trust they want to put in you vs the public CA infrastructure.