Hi, I'm not sure how to diagnose this, but since one of the most recent updates (within the last 3 months) I noticed that some of my LIFX devices are missing messages. I notice this specifically when triggering a good night routine that turns off all the lights in the house at once. There are always a couple LIFX devices that didn't get the message, and they're different each time. This has never happened before, so I thought I'd connect with the community here to see if anyone else is experiencing the same.
I'm assuming you are using the LIFX Color driver and have a fairly new hub with the at least fairly recent OS. I have about 20 of these.
I do use the Room Lighting app to control a lot of groups. You might give that a try. The next suggestion is probably just me not thinking, but I originally I had rules that were triggered by a virtual switch turning on and off...like say all the downstairs lights. If say my nighttime rule called the virtual switch, instead of "running that rule" from my nighttime routine you end up with multiple rules running at the same time.
I mean I can't prove that's how the hub handles it, but changing over to running the other rules actions instead of launching the rule actions from by flipping the virtual switch...I hardly have a miss with anything anymore. I do know that when you run the rule actions from another rule...the main routine (my nighttime rule) waits for those events to finish before proceeding next line in rule that called it.
I have people question why I use a delay command of about 10 seconds between major changes but I think it helps...also when available the "command retry" helps too. IMHO
Good luck. I hate when I have random failures. I mean the very definition of insanity is doing the exact same thing and getting/expecting different results. 
Thanks, I'll play around. My good night routine is pretty simple: I have an "ALL lights" group. I can try disabling "enable on/off optimization"... Otherwise my next step is to do what you said and build in delays into the rule for the LIFX devices.
I did replace my router last month so that may be a factor. Everything is set up exactly the same way as before, with Static IPs. Only difference is I have all my home automation devices (including all the LIFX) on a separate IoT network now. The new network is an ASUS IoT thing... I haven't looked too deep into it yet but it's supposedly optimized for compatibility and legacy devices, and to keep it away from your home network. I have not gone so far as putting them all on a separate subnet, but I plan to...
On a rare occasion groups of LIFX devices will stop responding for me.. Normally, I reboot my WiFi access points and this solves the problem for me..
It's been a long time since it happened to me..
I will say that the room lighting app is really nice because you have all kinds of options for forcing off (or on) lights that the system thinks are already off for instance.
It is specifically made for handling groups of lights
I didn’t use it for my first couple years on this platform but it recently started because it’s pretty efficient.
Interesting, I just looked at the Room Lighting app. I still use Groups and Scenes for the Zibgee grouping. Zigbee groups FTW. 
When doing large numbers of lights and switches I would see delays and random failures with larger groups of lights. I've not worked with WiFi lights but it does sound familiar.
With Zigbee I would get the popcorn effect and not every light would work all the time - similar to your issue. This was resolved with Zigbee grouping.
In my living room I have 30 MiLight track lights that are unique. I have an ESP32 that sits between HE and the lights. If I select all 30 lights and tell them do something it will get slower and slower until a bunch fail to respond. Luckily the ESP code has a grouping function so I can setup a group like I do with the Zigbee lights.
I also had my Zwave mesh get super slow and stop responding. In that case it turned out a box was leaning against a light switch and flooding the mesh with commands. It took a while to fail. Eventually all of the Zwave devices stopped responding because of the flood of traffic from the light switch.
Of course this is all anecdotal, but the behavior was close enough to point out. In a worst case scenario you could try updating half the lights, wait 10 seconds or so, then update the other half. It isn't ideal, but if it works reliably it might point to something else (perhaps a mesh issue?). This isn't a fix, just a troubleshooting step.
I'm only seeing this issue with my LIFX lights. All other zigbee bulbs are doing fine. Never had issues turning off ~30 lights at the same time until recently (or the very early days of Hubitat)... and it's only the WiFi bulbs now, so the issue is network related.
Add a wireshark port mirror on the switch with the HE, and capture the packets headed to the IPs of the LIFX devices - That's somewhat of the beauty of WiFi, is that it's a bit easier to get detailed packet traces, and see which "end" of the flow that's dropping the ball, or looking for retransmits, etc. - This very well could be a WiFi problem.
I also use the older "groups & scenes" app, and for large groups of LIFX WiFi bulbs, I ended up enabling command retiries, and adding in a metering delay of 100 msecs in the group definition.
YMMV of course.
I get it, I was just pointing out I've seen random slowness when the radio gets flooded. I would still try half the lights, wait, then the other half - or even thirds. If that works then start looking at where there could be congestion on the network.
Do you have a new neighbor that might be microwaving some popcorn when it fails? What are you using for the WiFi?
I'm using two ASUS RT-BE88U with AiMesh, wired via ethernet cable.
But I just noticed I hadn't enabled Ethernet Backhaul Mode, so let's see if that helps.
