Why do RTC components occasionally get disconnected?

I'm not sure how those Eclipse projects would have gotten disconnected.   There is an explicit "Team -> Disconnect" operation which would disconnect them, but the user probably would have remembered if they had done that operation.   I'll see if I can find out other things that might cause an Eclipse project to become disconnected from an RTC workspace, other than this explicit operation.  (If anyone else on the forum has some experience with that, please post a comment here).

But there is a fast/easy way to fix this ... just "share" those projects into Team Concert again.  In particular, run the "share" wizard, and the wizard will just reconnect the files in the sandbox back to the RTC workspace.  The "share" operation to reconnect the Eclipse project is smart about unresolved changes ... it will detect them, and they will show up in your Unresolved folder.

One thing to be careful about ... the wizard will ask you what component to share the project into ... make sure to select the original component. If you select some other component, it will think you want to share this directory into that other component, and create a change set that creates all those files in the other component.   But if you make that mistake, it's actually not a big deal ... discard that new share change set, "disconnect" that project, and then run the "share" wizard again, this time picking the right component.

 Another way is to Load again.  Before you do that, make sure you backup your sandbox in case it fails, although it always work for me.  Do the normal Load process, select the same directory you had, when you click on Finish, it will give you a warning:  "The selected folders collide with projects already in your local file area.  Please confirm which ones to re-load".   You OK it, then Cancel the Load process, everything will come back

Yes - I mean 'unloaded' although as you pointed out the error message (incorrectly IMHO) says disconnected. There was no intervention by the user to deliberately unload all eleven components in the stream.  In answer to Geoffrey's questions: -  Yes all the files were on disc.  We verified that by doing a difference with what was expected.  My workspace was unaffected.  - No other user or process could access his local workspace, it is on a lap-top.  - It takes many hours to reload because thousands of files with gigabytes of data became 'unloaded', and we have a relatively slow line here (30 Mbits/second)  - There were file changes locally which had no yet been checked in, which had to be 'reapplied'  (probably the wrong word again). 

We have the same problem occasionally and we are not able to identify the cause... It seems occurring when the sandboxes are located on a network driver (i..e, NAS). Any advice?