This paper considers the problem of optimizing the measurement timing for direct in-space thrust inference when the propulsive acceleration is known to act in the along-track direction and can be represented by a piecewise-constant function. A linear measurement model, derived in prior work, is used to relate changes in the spacecraft's initial state as well as magnitude of the average atmospheric drag and propulsive accelerations to deviations in the spacecraft's position in the orbit plane from a known reference trajectory. The linear measurement model allows the posterior covariance of the propulsive acceleration magnitudes to be analytically calculated given a prior covariance, the measurement covariance, and the measurement times. The measurement times can then be numerically optimized in order to minimize the posterior covariance of the propulsive acceleration magnitudes. Example scenarios considered in this work show that the posterior marginal standard deviation of the propulsive acceleration magnitudes can be reduced by approximately 20% purely through the optimization of a fixed number of measurement times. The primary benefit of this work is for data-limited scenarios, where the number of measurements may be limited, or scenarios with finite sensor resources.