I've been studying the gameshape patterns of my players over the last two seasons and I've noticed a cumulative effect of several weeks of playing time on the current week's GS update. I think it could be as many as 4-5 weeks that matter, or maybe even the entire season's minutes up until your current week. Which would make sense as players would tire out at the end of a season they played too many minutes in. I didn't quantitatively track this though, so I'm not sure about it.
In Italian forum we had a big discussion some seasons ago about a statement similar to yours to explain the strange thing about GS
At the time my idea was that that players's Gs go through some trends during the seasons,and the minutes played in the single week work to exalt a positive trend,and to limit a negative trend.With 60 minutes played in a week,you'll have the best effect on your game shape,as always come out from all the studies...when you go far from a decent range(40-80 minutes) both as single week's minutes(close to 0/144) and multiple weeks where you miss a decent range(multiple weeks close to 30/90 or worse),you'll change the trend toward a negative trend
More experienced users instead say that GS work like a rigged nut,and minutes work to change the probability of a positive or negative outcome(improvement or decrease in GS decimals).The cumulative effect would come from the fact that with a similar number of minutes played in the various weeks,the outcome will be often similar and the result will accumulate during weeks,but it could however happen sometime a strange result for a single week
The general opinion was that the second solution is probably the right solution because it's basically similar to the GE(minutes instead of skills)