Thanks, but going virtual chassis is not a solution for us.
We want to keep 2 distinct control planes for core (2 distinct EX4500, each one of them running its own junos, its own RE), as we had to deal with severe outage due to usage of VC in the core (it happened on this architecture before I took my job here, and people decided never to go for VC on the core anymore).
So core will stay that way, with 2 standalone EX4500. In near future we may replace EX4500 with QFX and go the MC-LAG road. But even in this configuration Juniper recommend to run rstp on MC-LAG(core side) and LAG (access side) for loop prevention/misconfigurations.
Access ex4200 are actually VC of EX4200. Each server is attached to 2 virtual chassis.
My point was that some server (very few) seem to have been configured with rstp, and I'm looking for confirmation of my undestanding.
yes, block on edge is a solution. But i'm taking responsability of this architecture that have been designed by someone else, and there's no block on edge anywhere. If I activate this option, all the misconfigured server will be isolated from the network if they run stp, and I find it to be quite radical in a production context.
I'm trying to check my understanding of the counters first. Can you confirm wether what I wrote about counter is correct or not?