I would begin my investigation looking at the vSwitch and port group configuration, then extend that to the physical switch ports that the host physical NICs are connected into - see if everything in that area is configured as you would expect and there’s no mismatch.
Hi I've just double checked this, the vSwitch and port group config is fine. The physical connection is fine too.
More details: When the VM losses internet connection, it still can ping the HOST where it resides but not to another host or to the default gateway.
Anything else that I should check?
It sounds like it might be ARP/RARP related, so could still be a physical switch issue - with your topology, or the switch port configuration - something along those lines.
I‘m moving this thread to the vNetwork area of the forum (I‘m a moderator on here which allows me to do that) so that others with more expertise can chime in...
Any pointers on what do I need to check?
Is your ESXi networking based on standard vSwitch or distributed vSwitch?
Is there any special configuration on the vSwitch failover order?
What is the physical switch port settings on the ESXi uplink connection?
Did you check the vmware.log file in time of VM startup?Please mark my comment as the Correct Answer if this solution resolved your problem
What I could think of, is that the physical switch ports to which the hosts are connected have spanning-tree enabled.
Ensure that these physical switch ports are configured with e.g. spanning-tree portfast, or as edge-ports (depending on the switch vendor).