Rudder

I was not able to reproduce that.

I install a rudder server 7.3, with a 7.3 node ; then updagraded to 8.0 and added a 8.0 node ; then upgraded to 8.1 and added a 8.1 node.
Then I factory-reset all the nodes, and the correctly appeared in pending nodes.
And if I try to accept them, I have the correct error message (even if not well styled).

The difference in dashboard and pending is because dashboard use a special case LDAP request and not the same logic than pending node screen.
We can make the two converge, at the price of perhaps lower perf if there is a lot of pending node (but we should do it for consistency nonetheless).

So, I believe that:
- either there is something special with Rudder 6.2 nodes in Rudder 8.1 (the problematic node were using very old rudder agent)
- or in each case, there was a problem with the inventory save in LDAP, which let things a bit broken in an interesting and complicated way to reproduce.