The concept of time has been on the mind lately, particularly in relation to uptime, pulse checks, execution time, and response time, which are typically viewed as health metrics. These metrics can also be seen as security metrics, as they can provide insight into the security of a system. The idea that timing can be used as a security metric is not new, as it has been used in the past to detect covert timing channels and behavioral baselining. Attackers have long understood that timing can carry information, and researchers have demonstrated timing side-channel attacks against large language models. However, most research has focused on how timing can be exploited, rather than how it can be used to defend against attacks. A recent survey of LLM-agent security identified temporal anomaly detection infrastructure as an open research gap, highlighting the need for more research in this area. The survey noted that current agent deployment frameworks do not support the behavioral baselines needed for temporal anomaly detection, and the discussion has largely focused on session-level behavior. The space between observable events, or the rhythm within a single execution, remains largely unexplored, and it is possible that time is another dimension of observability that has been overlooked. By measuring behavioral changes and entropy across inter-arrival intervals, it may be possible to use timing as a signal rather than noise to improve security. The author is exploring this concept in their own work and is curious to know who else is working on the defensive side of temporal behavior, particularly for agentic systems.