Also note that rather than going through Python and Astropy you can retrieve an MJD directly as calexp.getInfo().getVisitInfo().getDate().get(lsst.daf.base.DateTime.MJD) for TAI timescale or ...get(lsst.daf.base.DateTime.MJD, lsst.daf.base.DateTime.UTC) for UTC timescale.
To be pedantic, the MJD of mid-exposure varies from one side of the field to the other (by 1s according to the key numbers, or 0.98s from the youtube shutter video). For rapidly varying or moving objects this effect may be significant. Whether or not normal processing corrects for this, does the metadata include the traverse direction for the double-bladed shutter such that users can later correct for location in the focal plane? Otherwise successive exposures will see opposite edges leading and then trailing the mid-array mid-exposure MJD by 0.49s. Will MOPS include such a correction in astrometry submitted to MPC? How accurately will the mid-array mid-exposure MJD be measured aside from this mechanical constraint?