It's normal for the rack to make a knock-knock sound when the engine is off (my 94 does the same thing) since there's no power steering fluid being fed to the rack. It shouldn't happen when the engine is running and fluid is being pumped into the system.
Internal to the rack, I don't see that as being true for the linear slide surfaces and mating of the pinion gear and rack teeth, which are mechanically downstream of the fluid-powered forces and motion of the pinion shaft "motor". (See my comments below on noise masking with engine running.)
Deep dive:
It may sound counter-intuitive, but, if you think about it, any forces and take-up-slop movements on the
output side of the rack (internal main rack rod, rack mount bushings, tie rods, McPherson struts, control arms, knuckle) will be the same (or less) with the engine off than with the engine/power steering pump running. But (with engine off), forces and take-up-slop movements on
input side of the rack (steering column-to-rack coupling, steering column intermediate shaft coupling, and steering column universal coupling) will be greater (because, without power assist, the input side sees huge - almost brick-wall - resistance to movement on the output side).
What may fool you into thinking otherwise is that with engine running, the noises of the engine, power steering pump and fluid, and actual turning of the wheels (direction-wise - i.e., tire tread contact patch rotating on pavement, concrete, or gravel) may be masking any take-up slack noises resulting from back-and-forth steering wheel motion that would be present with or without engine running. If what I'm saying doesn't seem right on first-time-thru mental visualization, think it thru a couple of times to see if it doesn't start making sense.
Bottom line:
The key is to realize that, for a given back-and-forth motion of the steering wheel with the engine running, the forces on all components on the the outout side of the rack will be greater, and the forces on all of the components on the input side of the rack will be lower than with the engine not running. Resulting take-up slop noises on the output side of the rack may only
seem to be less with engine running because the additional noises (engine, etc.) will be masking the take-up slop noises - you'll just have to listen "harder" for them (or go more by visual than by sound). And don't rule out that take-up slop/knocking noises on input side of rack may be what you're hearing (especially with engine off). Input-side noises should/would, in reality,
and because of masking, be way less or non-existent with engine running.