Stack member loss should generate a "warning" severity event and corresponding SNMP trap. Same for removal/insertion of the stacking module itself. The trap isn't generated by default, you have to configure your trap receiver for a trap-level of "not-info" to receive it. Agree this should be generated by default.
snmp-server host 128.44.120.99 community "public" trap-level not-info
W 04/26/16 17:34:26 03258 stacking: ST1-CMDR: Standby switch with Member ID 3
removed due to loss of communication
I 04/26/16 17:34:25 02559 chassis: ST4-MMBR: Stack port 1 is now off-line.
I 04/26/16 17:34:26 02559 chassis: ST2-MMBR: Stack port 3 is now off-line.
W 04/26/16 17:34:26 03270 stacking: ST1-CMDR: Topology is a Chain
2016-04-26 14:59:30 128.44.112.100(via UDP: [128.44.112.100]:161->[128.44.120.99]:162) TRAP, SNMP v1, community public
SNMPv2-SMI::enterprises.11.2.3.7.8.5.2 Enterprise Specific Trap (2) Uptime: 7:56:28.00
RMON-MIB::eventDescription.3258 = STRING: W 04/26/16 17:34:26 03258 stacking: ST1-CMDR: Standby switch with Member ID 3
removed due to loss of communication