Network error on Confluence login

[CCCP Admins]

I've a server on which both Confluence and Jira are running. When a registrated user tries to login on Confluence, it gives the following error:

Sorry, a network error occurred trying to log you in. Please try again, and contact your administrator if the problem persists.

Confluence's error log gives the following 'explanation':

2012-04-25 20:52:12,782 ERROR [http-8080-2] [crowd.manager.application.ApplicationServiceGeneric] authenticateUser Directory 'Crowd Repository' is not functional during authenti$
2012-04-25 20:52:12,783 WARN [http-8080-2] [atlassian.confluence.user.ConfluenceAuthenticator] authenticate CommunicationException caught while authenticating user <bob>
com.atlassian.crowd.exception.runtime.CommunicationException
[...]
Caused by: java.net.ConnectException: Connection refused

each time a person tries to enter the confluence wiki on my server. The userdata is retrieved from a LDAP database.
This whole problem appears to come out of the blue: this one moment everything was working, the next Confluence decided to gives this anoying error.

[CCCP Admins]

Crowd was indeed the problem. Upgrading crowd from version 2.2.4 tot 2.4.0 helped to solve the problem. Probably a file (or multiple files) of Crowd became corrupted for unknown reason.

[Thomas Krug]

The error shows you are using Crowd. Does Crowd contact the LDAP then? Can you reach the Crowd server?
Even though I may sound queer, did you try and restart Confluence?

Errors comming out of the blue are often results of inconsistent caches, buffers etc.

[CCCP Admins]

I'm indeed using Crowd to make contact with the LDAP and it is installed on the same server as Confluence. The LDAP however is located on another server, but Crowd would be the only program to have trouble with the connection, if this is the source of the error.
Restarting Confluence is a good idea, but I've thought of that already and have restarted it (many times), just like I've restarted Crowd (also many times).

[CCCP Admins]

Research of the Crowd errorlogs shows that Crowd didn't produce any errorlogs since the time Confluence gave the error. For this, I think Crowd is the problem.
Well, checking the processes on the server, I've concluded that Crowd won't run at all. Everytime Crowd restarted, it did essentially nothing. Appearantly Crowd is so broken, it won't even say anymore it IS broken. Question remains: why is Crowd broken? Or more important: how can i fix it?