This was discussed 8 years ago. Wow :-)
The content of that older thread I think remains fully valid.
My question just goes a little further than the earlier discussion did:
If I set up the same DHCP scope on two co-operating RB1200s in a VRRP configuration, then I imagine that some clients will get answered by one RB1200 and other clients will get answered by the other. Each RB1200 will think, for any IP address in the same DHCP IP pool that it DIDN'T server, that the address is simply "in use". We won't run out of addresses, because only one RB1200 will actually server each client at any one time.
Here's the more detailed question, then:
When a failure occurs, and ~half of the clients have an IP address from the failed RB1200, and then their DHCP lease renewal time arrives, they will be unable to reach the RB1200 that served them. They will then send a more general DHCP request, which only the other RB1200 (which thinks that their current IP address is in use NOT served by its own DHCP server) ... do we then start running out of IP addresses, because the remaining RB1200 knows that the IP address which the client prefers "is in use"? .. Or is DHCP and the MikroTik RouterOS 5.x implementation of DHCP smart enough to recognise the fact that the requesting client IS the node which already has that IP address in use, so the RB1200 will change that IP address' status in the RouterOS DHCP tables from "in use" to "assigned BY ME to this client"?