While adjusting the hold timer is possible, it is strongly not recommended - see the sk44711 Kernel debug shows that TCP traffic is "dropped by fwhold_expires Reason: held chain expired" for detailed explanations. The hold table change ic covered in sk107324 Some traffic does not pass through Security Gateway and kernel debug shows "...dropped by fw_filter_chain Reason: chain hold".
There is no calculation for the optimal size - you have to raise it slowly until the drops disappear. Memory consumtion will grow of course...
CCSP - CCSE / CCTE / CTPS / CCME / CCSM Elite / SMB Specialist