We present a list of obstacles as well as possible solutions for doing multi-wavelength time domain science on large scales inside of science platforms. Relevant to this talk, science platforms are computing environments provided by archives near the data which allow fast, convenient data access and computing, which thereby increase inclusion and reproducibility in science. Our specific use case is to generate light curves from many available archives at many wavelengths for a sample of 500,000 quasars. In writing this use case, we have hit stumbling blocks at 1) determining the best data structures to store and work with time-domain data, 2) finding the best way for archives to serve large time-domain catalogs so that scientists can 3) access those catalogs, and 4) concerns about understanding the calibration of large multi-wavelength surveys. For each of these obstacles, we discuss our requirements as well as solutions we have developed to address those obstacles at scale.