All tallies will give you quantities that are “per source particle”. Since each source particle is lost (absorbed or leaked), this means that “per source particle” is really the same thing as “over the total loss rate”. I ran a quick test on the Jezebel benchmark using OpenMC and MCNP and get the following results:
MCNP prompt removal lifetime = 3.7435e-9 +/- 2.7119e-12
OpenMC inverse-velocity = 3.73529e-9 +/- 3.84415e-12