These are some of the worst. If you're able to get them on to a different network--they work fine with a Hue Bridge, for example; or some use a second Hubitat--this would be a way to rule that out. (I actually tried the "dedicated Hubitat for Zigbee bulbs" route and they managed to fall off that network, but others have reported them being fine; I've had no problems with them on Hue.)
It is interesting that the Sengled is the only device you notice this problem with, but my guess there would be that the Sengled always chooses the Cree as its route because it perceives one of those as an attractive repeater from whatever algorithm Zigbee uses to compute this. You can't control Zigbee routing, so eliminating the Crees from the network is the only way to be sure, but you can see limited route information from an undocumented Zigbee tool at http://YourHubIP/hub/zigbee/getChildAndRouteInfo (not much explanation but some here: Zigbee Child and Route Info Table Help / Explanation). You don't always see all devices in that table, but if you can tell that a Sengled is routing through a Cree, that would at least provide more evidence for this guess. But if you could live without those bulbs for a while (or long-term move them to a different network), that's probably best regardless.
