Any operational network has some excess capacity (EC) to avoid early exhaustion of resources. We propose to exploit the EC in optical WDM backbone networks to support efficient and survivable traffic grooming where connection requests are of sub-wavelength granularity and each provisioned request has to be protected from a single link failure. Our novel EC management techniques can improve network performance, at no additional cost to the operator, since excess capacity is normally unutilized. Our techniques exploit EC such that a connection can use a protection scheme which provides high reliability but may consume more resources when traffic is low, but it switches to another protection scheme which provides lower reliability but is resource efficient by reprovisioning backup resources. As a complement, we propose hold-p-lightpath scheme to exploit EC by preventing the termination of pre-established (but unused) resources. The backup reprovisioning problem is split it into three subproblems: when, how, and what to reprovision; and we propose our solutions for each subproblem. For the what to reprovision subproblem, we design three methods with different reliability and resource-efficiency performance. We compare our approaches with traditional protection schemes for typical daily fluctuating traffic, and show that significant improvements in performance and cost can be achieved.