Over a large enough sample size a player will eventually revert to his mean performance. Sadly for us, one season (especially considering all the variables) is not a big enough sample size. Regardless, it is a big enough that extremely outlying performances are rare. For instance, I am in a 95M league right now... even taking into consideration that most players at that cap will underperform... having my .371 BA (RL) Tony Gwynn batting under .250 for the first 30 games was a concern. The season is pretty much over now... Gwynn got "hot" and will finish the season at approx. .315. I'm not disappointed. In the SIM, patience is usually a virtue.
Ideally, when you drafted your team you made the best objective decisions you could... in an OL, any player was theoretically available to you, and you chose who you chose. If at some point you decide you don't like a guy, going to the WW just means that, with the 10% loss in salary (and the smaller number of available players) you are probably going to end up making your team worse... especially if you do it a number of times. I've seen some OL teams WW their way out of nearly 20M in salary over the course of a season. Regardless, I can see three reasons for using the WW anyway... 1) you're new and didn't factor in things like normalization, so your 1930 BA guy or your late '90's power guy is never going to perform to the role you selected him for 2) you're new and you picked the wrong type of player for your stadium or you were very unlucky and the stadiums of your direct competitors have combined to negate you player's main strength 3) you got lucky with your AAA guys and can dump some of your bench players, get some extra $$$, and actually upgrade at a position.
Sosa will come around. His bad start will probably drag down his overall numbers, but it is probable that he will eventually get "hot". Keep in mind though, hitters from the mid-90's to the mid '00's will underperform compared to their RL numbers. Power hitters especially, HR-wise.