Hi,
As the saying goes 'The devil is in the detail', this might well be the case.
If the inet addresses are being configured under the physical ge interfaces on the MX, vrrp pkts will be sourced from that interface and unless the SRX is bridging/switching the vrrp packets, those will not reach the other MX and vice-versa, and result in split-brain vrrp as the back to back L2 trunk interface between the MXs will not be used for vrrp control traffic.
From experience, I believe the MXs can be configured with a bridge-domain and the vrrp configured over an irb. That way vrrp can flow over the L2 trunk. Also to make sure data traffic does not traverse over the L2 trunk, vrrp track and priorities can be used to ensure vrrp mastership is in sync with the active reth on the SRX side.
Hope this helps.
Ashvin