@fractal_boy Sure but thats a bit legacy when it comes to datacenter design to be honest.
Having multiple VRRP groups isn't a modern design. VARP also wont have traffic run over the MLAG link but VRRP will.
Even thinking about multiple VRRP groups in my use case isn't scalable. I have over 100 SVIs in production and development environments. Its starts getting unwieldy
edit:
Looking at your drawing, forget end user or servers downstream. Pretend its a dual homed switch or server that has a LAG up to the MC-LAG. It can send frames to either spine. That alone would warrant VARP.