What does "Good Enough" mean, I'm unclear on what "success criteria" your using here.
It makes sense, that the NTP clock set (as the HE has no battery backed RTC), occurs on boot, and that clock drift/skew is fairly small and would take days/weeks to accumulate to meaningful time errors.
So if it boots, connected to the network, it gets an IP via DHCP (unless you have a static IP set - NOT a reserved address), and NTP initially sets, the clock - You can then unplug the e-net cable, and it will locally run along just fine doing Zwave/Zigbee things (matter will fail, as it's IP based, and any "offsite" stuff will fail as you noted).
If it's NOT plugged in (or no network access on boot), it will NOT get a DHCP address, not will it initially set it's clock correctly. - The router may still have a DHCP lease cached, until the lease time expires (as there is nothing on the network to reply with a NAK). With a static IP, you can work around the DHCP failure, and it should at least boot cleanly, but without setting the clock via NTP, no time based rules are going to work well (clock defaults to Jan 1, 1970 which is the UNIX epoch? - That's a SWAG on my behalf) - I would assume Z-Wave/Zigbee triggers on buttons/motion sensors/contacts/rules will still work fine without the clock being accurate. Obviously with no local GUI or web access.
But again, back to the beginning here - I'm unclear on your goal, everything above seems to make sense, given the lack of any IP access/connectivity, but what do you want to have happen to declare "success" or "good enough"?
Or you can leave it on your local network, an isolate it via a VLAN, and control what can talk to HE and vice versa via ACL's or FW rules - But that's all pretty much standard network security control stuff.
Or is there something in this thread that is unexpected to you, and your trying to work around some issue?