This fixes a bug affecting the smearing step of jet calibration for jets with eta values very close to zero (roughly |eta| < 1e-5). These jets would result in passing an argument of 0 to TRandom3, which is a special value resulting in nondeterministic behavior. This simply adds an offset to the seed so that this can't happen anymore.
Note that this will change the output for smearing of all jets with respect to previous versions. This is fine, but anyone doing precise validation checks, CI tests, etc. should be aware of it.
This has already been fixed in master, so no sweep is needed.
Tagging @sschramm, @cdelitzs, @sawyer for their info - we should probably make analysis groups at large aware of this, it came up because a SUSY analysis noticed the effect in one of their CI tests.