The stories of people having service issues is not desirable, but also largely unavoidable. Using non-custodial wallets likely won't change the fact that there are issues either, just change the kind of issues that people have. That said, its never good to not have control over fixing the problem because it relies on someone else to fix it - but all the same, anyone new isn't going to have any tools to do it either, and likely won't feel like its any better to have some stuck thing in the wallet with or without their keys.
Luckily Phoenix and Breez have been incredible and I've not had an issue in a really long time. Probably getting to a place soon where so much is ironed out that the most common of those sorts of things fade away.
------------------------
On the Chart
Not really, because this doesn't delineate the type of "usage" and I suspect its sending and receiving, not balance related.
In other words - this is exactly what I would expect with the rise of zaps and the use of lightning addresses (since they are almost always going to be custodial), and it doesn't tell us anything about the funds at risk, because this (I assume) is measuring *payments.*
IF this was showing lightning *balances* with this trend I'd be a little more concerned and would probably fuss about getting BOLT12 at twice my current rate. But the funny thing is that I'm 100% contributing to this chart trend, but it's not because I'm putting all my money on custodial lightning.
I receive all of my zaps either from WOS, Alby, Fountain, or LN.tips and send about half of them from the same wallets (when I use Damus its all phoenix). So custodial traffic on my custodial wallets is probably 10x the traffic of my non-custodial wallets. However, whenever I get more than a few 100 thousand sats I withdraw to phoenix or my main node... which may only end up being a few transactions, but it is obviously the overwhelming majority of my BTC that I use on lightning.
-----------------
TL;DR - The chart appears to show number of payments, not amount of BTC at risk. With zaps I would expect it to look exactly like this but my major concern is how much people have at risk, not how many individual zaps they send or receive with non-custodial wallets vs custodial.