4a) The problem would be that if a group of people who work together all get the same treatment, they might all respond in a similar fashion, because they all might have fun together and then you'll think that that treatment lowered stress when in fact it was the fun they were having together. Random assignment of treatment will balance out any similarities these people might have with each other and help create homogeneous treatment groups for a more clear view of how the treatment effected the volunteers.
4b) A control group gives us a baseline for comparison. Something might happen during the course of the study, say, everyone gets a raise. Then we could see how the stress level was reduced overall and will be able to tell if the treatment(s) lowered the stress even more.
4c) No, the study is done on volunteers, not a random sample of the company. There might be differences from the volunteers compared the average worker.
