I noticed descrepancies like this a while ago. All the data around the 8/31 you mention is off by similar amounts.
One guess is it might have something to do with how each data vendor deals with the ex-div date price adjustment/reduction. For instance look at June 15. That "should" have been a gap up, and stockcharts.com shows it as a gap up, but TOS shows the ex-div adjusted data so there is no gap up and instead June 15 opens right around the prior days close. And yeah SPY does have it's share of bad ticks and some data vendors never corrrect them, so that could be part of the problem too.
A while back I compared Thinkorswim, Prophet, stockcharts, freestockcharts and there were definitely days where SPY didn't match up across the four. But from the stuff I looked at $SPX always matched up to the penny on highs and lows so I'd say that's more reliable, even though it sucks that it doesn't show the gaps and always shows the open = prior days close.
EDIT - what's interesting is although TOS and stockcharts.com show different price action on the 6/15 ex-div date, the highs/lows match up since then. But PRIOR to the 6/15 date, sc high/lows don't match TOS. So it looks like sc is adjusting high/low data prior to the ex-div date, to account for the fact that they don't show the ex-div price reduction. Maybe this is the reason?
But there are a lot of other smaller descrepancies if you look at enough data, probably due to bad ticks. So to be safe I stick with $SPX as my primary source for S&P levels but still look at SPY to get a better feel for the gaps, etc