Strange disservices with FireEye and FortiGate

Hello to all engineers, I would like to submit to you a rather particular situation to which I cannot find a solution.

I have a rather small network where in site A there is a LAN / 24 populated by physical PCs and VDI; the VDI reside in site B, which is equally large.
Site A is connected to Site B via a circuit managed by a 100 Mbps provider (which never saturates) that talks to an MPLS network to which a FireEye device is connected and immediately after it there is another firewall.

Site A and Site B firewalls are FortiGate v6.2.7 and do not suffer from any particular problems.
Every Wednesday morning, between 9 and 12, there are strong increases in response times between offices A and B, so much so that Teams calls and VDI clients freeze and then disconnect.

The FireEye is not in my management and the technician is investigating any scheduled jobs and errors, but until now he warns that everything is ok.
Randomness, however, wants FireEye to be connected only to the Master node of the firewall of location B and when there are slowdowns, I switch to the Slave node and the problem is solved, everything returning to normal.

If there is no saturation, RAM / CPU overload of the network devices, but if switching on the node to which FireEye is not connected the problem is solved, what can be the reason for the inefficiencies? What analyzes can I carry out? How would you behave?
The problem often happens on a Wednesday, but it is not mandatory, it may not happen again for a week or two.

I can't reproduce it on command.

Thank you associates!