We're running DHCP on five servers (6.5.5) at different WAN locations. Each
has a complete read/write replica of NDS. It seems whenever a WAN link
drops and the a server can't communicate with the others, DHCP services
fail at load. I can't confirm yet if running DHCP service also fails. Staff
at these sites are pretty hasty to reboot servers.

Is a read/write replica of the local DHCP objects adequate, or do I need a
local master partition?

Could this be affected by loss of timesync?

Could this be affected by loss of SLP communication?

Any other guesses?