It only happens when high load of traffic ?
Then the failure is probably linked to a board congestion. It’s not really a HW failure. You’re just outside the nominal use of the boards.
The board cannot handle more calls and drops on-going ones. Or something like this.
Could you compare your traffic carried by the whole BSC at BH with its capacity (check the BSC configuration, number of CCP boards, or TCU / DTC boards, etc)