Make the internal kpackage job task less fragile
This is done to ensure we don't end up emitting things on an already-deleted object, which breaks. We cannot simply swap the existing code out for a connection that's queued, as we may end up not emitting the result, which would be a problem in itself.
We also put things on the heap instead of the stack now, and move a few internals around a bit, so we don't end up deleting things we need.
Also clean up the error reporting a touch while we're at it.
Edited by Dan Leinir Turthra Jensen