The only real issue that I faced was while modifying the access list. That's the only time the outage is likely to occur, ie when the access-list is removed and a modified one re-entered. What we tried is to modify the access-list while still in place. We used the no form of the command as shown below
ip access-list extended 101
no permit ip host
host We then moved in the new statements (to the bottom of the list by default). The access-list got modified sucessfully. This was attempted in the lab, and although on the production network we opted for downtime, there is no reason why the same can't be done on the live network. By doing this removing the crypto map is not necessary any more.