Industrial applications require more and more low-power operations , low-delay, deterministic communications as well as end-to-end reliability close to 100%. However, traditional radio technologies are sensitive to external interference, which degrades the reliability and introduces unpredictable delays due to collision detection and retransmissions. Therefore, recent standardization efforts focus on slow channel hopping strategies to provide strict Quality of Service (QoS) for the Industrial Internet of Things (IIoT). By keeping nodes time-synchronized and by employing a channel hopping approach, IEEE 802.15.4-TSCH (Time-Slotted Channel Hoping) aims at providing high-level network reliability. However, some radio channels still suffer from high external interference and need to be blacklisted. Since the interference pattern is rather dynamic, unpredictable and highly localized, we here propose heuristics to decide which channels to blacklist. To avoid deafness, the transmitter and the receiver must also agree on a consistent blacklist. Furthermore, since the external interference may be time-dependent as well, we also propose mechanisms to decide when a channel has to be blacklisted or on the contrary recovered. Our thorough experimental evaluation based on OpenWSN and FIT IoT-LAB highlight the relevance of this approach: with a localized blacklisting strategy, we increase by 20% packet delivery rate for the worst links.