Combustion facilities run on pulverized biomass often exhibit fluctuations in fuel feeding and, thus, equivalence ratio and would benefit from fast process control based on optical lambda sensors installed in the reactor core. The conversion of softwood powder is investigated in an atmospheric entrained-flow reactor (EFR) operated close to stoichiometry using two different burners (swirl and jet) and three oxygen concentrations (21, 30, and 40%). Tunable diode laser absorption spectroscopy (TDLAS) is used to conduct time-resolved (0.1-1 s) in situ measurements of the gas temperature, carbon monoxide (CO), water vapor (H2O), gaseous atomic potassium [K(g)], and soot volume fraction in the lower part of the reactor core and in the exhaust of the EFR At both locations, the measurement parameters show significant, correlating fluctuations. The local equivalence ratio is derived from a comparison of measured CO and H2O concentrations (for fuel-rich and fuel-lean conditions, respectively) to thermodynamic equilibrium calculations (TEC) and found to vary in a wide range (0.8-1.3). Soot production decreases with an increasing local equivalence ratio and oxygen enrichment and is lower for the swirl compared to the jet burner. The measured K(g) concentrations follow the general behavior predicted by TEC around stoichiometry. In the relevant temperature range (1100-1700 K), K(g) is 2-4 orders of magnitude higher under fuel-rich than fuel-lean conditions, with a sharp transition at stoichiometry. While K(g) concentrations are lower than TEC in the reactor core and under fuel-rich conditions, excellent agreement is found at the exhaust after complete fuel conversion. Precise, wide dynamic range detection of K(g) using TDLAS enables discrimination between fuel-rich and fuel-lean conditions and has the potential for lambda sensing close to the hot reaction zone of combustion plants.