Due to the special requirements from the customer, JTAC worked with PLM and the engineer team to review the architecture.

The Juniper documentation discusses EVPN Multihoming only from the perspective of a single CE device multihomed to one or more PE routers where the routers provide an ESI all-active LAG.

https://www.juniper.net/documentation/us/en/software/junos/evpn-vxlan/topics/concept/evpn-bgp-multihoming-overview.html

single CE router ===== { PE1, PE2 }

We have the following possible scenario where we would have an ESI all-active to ESI all-active LAG.

On the Fabric side these links would be ethernet switching and on the MX/PE side they would be bridge domain in a virtual switch instance.

{QFX1, QFX2} EVPN AA ====== EVPN AA {MX1, MX2}

Based on the RFC document – RFC 7432, Section 14.1.2

https://www.rfc-editor.org/rfc/rfc7432.html#page-42

14.1.2.  All-Active Redundancy Mode

[…]

Consider a CE (CE1) that is dual-homed to two PEs (PE1 and PE2) on a LAG interface (ES1).
 

JTAC noticed that there is supposed to be only one CE in the standard topology, which connects two PEs in MPLS EVPN all-active redundancy mode.

The design had been reviewed by the PLM and the engineer team, and based on the best effort, it should be good to go.