In your startup script for each box that HACMP runs on, have it automatically adjust those parms, have it run a file with all the parms you want to change. Then have HACMP stop/start the db 1 more time before you have what ever connecting to it.
failover
dbstart
db2 udpate etc;
db2stop
db2start
Should only add about 1-2 mins at most to your failover.
The override bufferpool would work, but it limits all your BP, so if you have more than 1 it would limit that too. That parm is normallly used to fix an over allocation of memory, not to run on.