In industrial production, people, materials, tools, waste, and packaging interact and move. We formalize this in terms of flows: flows of people, as well as materials, packaging and waste, etc. A flow can be visually represented as a so-called spaghetti diagram. We say that a flow is regular if it does not violate the relevant production norms. Otherwise, we say the flow is irregular. We focus on analyzing paths between different flows because those may violate the production norms. A key goal is to prevent risky paths. Whenever the quality assurance procedures for production happen to exhibit violations of the relevant norms, this triggers recovery procedures to enable a swift return to regular flows. However, such recovery also eliminates the forensic evidence which would enable the identification of the cause of the production problem. Therefore, it is proposed to equip people, containers for materials, tools, etc., with Ultra High Frequency (UHF) Radio Frequency Identification (RFID) tags to enable their tracking of positions. Sufficiently accurate position tracking provides the data needed for the statistical inference about regular and irregular flows. These inferred flows can be visualized as spaghetti diagrams and may well serve as forensic evidence for post-mortem cause identification after the occurrence of a production problem. We report on position tracking from UHF RFID received phase and signal strength in an experimental lab scenario. The lab scenario consists of a single UHF RFID reader, which is connected to a distributed antenna system installed in a lab room. The antennas are multiplexed in time with remote-controlled phase-stable coaxial switches. The experimental lab scenario provides detailed maps of received signal strength and phase for all tag positions in a rectangular area in this lab scenario. We report on statistics for tag localization, tracking, and inference on trajectories For the realized setup based on 11 reader antennas, 99.96 percent of tag positions are covered by at least four reader antennas which indicates the feasibility of high precision localization at 0.04 percent outage.