So there is only 1 single change you made, i.e. MAT_RIGID to MAT_PLASTIC_KINEMATIC ==> and dt drops?
Can you check to see which elm type control dt? LAG or ALE? I think SW2 may give that info.
Are you using ALE or SALE approach?
Did you verify the mat data accuracy?
Same rho and bulk modulus used for both mat types?
Did you see any leakage in either model? If there are any, you must first correct that, by defining and tuning PFAC load curve.
dt_max ~ dl_min/C
dl_min = min elm width
C = sound speed of that medium
Note: modeling environmental air space as vacuum may make it easier for the coupling to leak. Using air provide a back P of 1 atm may help stabilize the coupling sometimes.
Those are the things you can check.