The fundamental condition for the legal approval of dynamic spectrum access approaches is the protection of the primary user. However, for dynamic spectrum access to become an attractive service reality, it is crucial to also ensure some quality of service support for the secondary user communication. In this article we discuss sensing-based opportunistic spectrum access approaches, in which primary user protection is achieved by a properly organized sensing process and secondary user communication reconfiguration. While the required reliability of the sensing process can be expressed in terms of rarely enough overlooking the primary user, we assume that the proper QoS for the secondary user is given by maintaining - with a given confidence level - a minimum bandwidth availability for the secondary user in spite of primary user dynamics. In this article we present an overview of approaches that might be used to achieve these objectives. In addition, we point out that both the sensing process and the secondary link maintenance (necessary to keep the required bandwidth in spite of reconfiguration due to detected primary users) require significant spectrum overhead. We identify and elaborate a fundamental trade-off in using these overheads in either sensing or link maintenance, and present examples of its optimization.