KGW won't do anything meaningful for Bitcoin. Even if the Bitcoin difficulty is adjujsted every block, there will still be periods where the average time between blocks is more than 10 minutes. Since the process of finding a block remains a Poisson process, regardless of what algorithm you use to tune the difficulty, block-delays of 30 minutes or more remain a completely real possibility.
The KGW algorithm is designed to combat a block-drought that comes from a sharply decreasing hashrate. As long as the hashrate doesn't go down rapidly, there is no real advantage to using KGW. If blocks take 11 minutes on average instead of 10 for a week or two, then that hardly matters and people won't even notice it unless they look at the data, since the variance in block-delays will obscure these structural deviations from the target average.
And since there is no reason to assume that the Bitcoin hashrate is going to sharply decrease (and not recover), there is no need to update the difficulty algorithm, which would require a hard fork.
Certainly there will be variability between block times under any scenario. However recently with the rapid increases in hashing power block times have been averaging as little as 7 minutes. This slightly accelerates the bitcoin inflation ahead of schedule. While I agree bitcoin is unlikely to have a sharp decrease in rate without recovery we are experiencing a sharp decrease right now.
https://blockchain.info/charts/hash-rateAnd right after a major jump in difficulty.
https://blockchain.info/charts/difficultyIt's a better system and while it won't serve much use during constant increases in mining power nothing goes up forever, especially exponentially.
The hashrate shown by blockchain.info (and other websites) is not the actual network hashrate, but rather a derived quantity calculated from the time between blocks. If the network hits a streak of bad luck and only solves 3 blocks in 1.5 hours or so, this'll show on these types of graphs as a drop in the network hashrate.
In reality, I assume that the network hashrate is rather stable, with some additional power being added on constantly. A drop in true hashrate only occurs when mining hardware is switched off and outside of catastrophic outages in datacenters, I don't see that happening on a large scale.
And as long as there is no sudden, large and non-temporary drop in the hashrate, the KGW algorithm doesn't really add that much. The downside of switching difficulty algorithm is obvious: It requires a hard fork. Hard forks are a big deal. And that simply won't happen for a matter as insignificant as this.