"Knocked loose" was NOT introduced all at once at a mystery time mid-season a couple seasons back. BUT, some things around it were changed that caused it to increase in frequency right in the same season that they added "knocked loose" to the pbp text.
If you want proof, go back and look at old games pre "knocked loose" text and add up the catches by your WRs and the completions by the QBs - they often don't match, and that's because of a stat bug that caused the QB to still get a completion recorded when the ball was caught but knocked loose - however the WR didn't get a catch (or a drop - that was added later too).
However, it obviously uses the FFumble check for knocked loose, and FFumble *was* tweaked about then. Add into the fact that it was right at the time of the ISlam debacle where all the defenses were pouring on Str trying to stop the falling forward for yards and broken tackles, and there you have it. In fact, when the FFumble was first tweaked (silently, I believe) the deflection on the ball was so great that the balls knocked loose were getting intercepted at a horrific rate, leading to several weeks of knocked loose - interceptions skyrocketing. So Bort turned down the amount of upward deflection on a knocked loose and those got back under control.
So here we are - currently WRs and TEs probably have more KLs than they should. And it's the age old question - tweak the sim or let the builds adjust? It's hard, because the pro level builds have gotten to a point where they are stagnating and the pros won't be different via builds for several more seasons. At the same time, we all know how awful it is to tweak the sim when builds will likely adjust to address this (DEs were so terrible for so long, and it really didn't need to happen as OTs were finally catching up build-wise).
Honestly, this is where testers could possibly come in handy, but not without letting them test out various builds specifically to see if certain amounts of stats create reasonable players, which is exactly what people don't want to happen on the test server as the data isn't likely to get disseminated well. WRs are so expensive that there aren't going to be enough experimental builds running around, either. It'd be nice if some people would mock up some various reasonable build choices (not completely maxxing out the stats all the way, but just pretty close to the right #s you could get) and see what different realistic choices you have for 50s and 60s WRs, and then have the testers play around with those builds and see what the tradeoffs are with each build. If you have to choose between blazing fast but gets jarred loose frequently and slower but able to bring down the ball in traffic and hold on when a LB hits him, then that's an interesting trade-off. If instead you have to choose between fast but can't pull in a catch, or able to hold on but not able to break out of coverage to get the ball, then it's a crappy trade-off.