Sever Push Problem with unreachable UIs (Request is not longer valid)

Hello,
we’ve got a problem with UIs which are no longer reachable/seen by the client.
When someone loads our application, they get a websocket connection, and everything is fine.
As there are some components in the UI we’d like to update later, we save these into a global state.
When the User reloads the page (Ctrl+F5), the user gets another UI, which does the same.
When we later update the components and push the update to the respective UIs, we’ve got the problem, that the first UI is no longer reachable by Websockets, as the client closed the connection during the reloading of the page.
Vaadin doesn’t know/care about this though, so it tries to send the updated data to the old UI anyways.
Atmosphere then detects that the AtmosphereResource is not in scope (atmosphere terminology), aka the websockets connection has been closed, and posts the following error to the console:
WARN | 13:16:15,902 | -Shared-DispatchOp-8 | org.atmosphere.cpr.DefaultBroadcaster | Request is not longer valid

Is there anything we can do about this? Ideally, the UI would get destroyed, as it is no longer rechable anyways. As we only hold it weak, this would also lead to us not doing anything on the UI, which would save us useless work, and the atmosphere warning later.
Any ideas?
Regards,
Cornelius

Hi, this is something we’ve been seeing intermittently as well. Vaadin registers listeners (AtmosphereResourceEventListener) that should be notified of client disconnections and clean up the server side accordingly, but for some reason or other they don’t seem to be completely reliable.

Is there a bug report for tracking this?
Or should I open one?
Btw Atmosphere definitly knows that the AtmosphereResource is no longer valid, as it’s inScope member variable of the AtmosphereResourceImpl is false.

Please open a ticket; there doesn’t seem to be any that’s specifically about this, although a couple might be related or caused by it.