I looked at this a little further.
From '87 to '89, the correlation between ES goal differential and NOMS was: -57%, -32% and +15%. So the correlation doesn't seem to be due to the direct effect of net NOMS on team performance. If you divide the teams into two groups, the extremes and moderates by ESGD, the difference only becomes more stark. For instance in '87, the extreme 10 (top 5 and bottom 5) had a -77% correlation, while the extreme 6 (top 3 and bottom 3) had a -92% correlation. Meanwhile, the moderate 11 had a +21% correlation and the moderate 15 had a +4% correlation. These are the types of numbers one sees league-wide in '89 (+15%). I also had data from the past 5 seasons, so calculated that and the correlations averaged +11% with a range of 0% to +22% during the 5 years.
It's apparent that the league must have been favoring weaker teams at the expense of better teams. Why and how they suddenly changed within a couple seasons is difficult to say. Perhaps the increase in PPs caused the refs' focus to go from evening things up to calling more penalties in general, but the change is fast and furious. Another factor, suggested by the data, is that as more penalties were called, and called more fairly, teams became more conscious of avoiding penalties against strong PP teams. They could get away with things earlier, because the refs weren't calling so many penalties, particularly against weaker teams. Once this changed, teams' strategies likely also changed.