The question doesn't really make sense this way. Thread is more or less like Wi-Fi. It's a transport technology and protocol. It's HOW your devices can talk to each other. It's how they can say "I'm here and I can see 'AirPurifier324' over there."
Matter is a bit like HTTP. It's WHAT your devices say to each other. It's a way for them to say "Hi, I'm a lightbulb and you can change my color and brightness."
Can it operate without an internet connection and with an open standard that lets the me, the user, to be in control using a hub (if necessary) and software I choose, including an open source option should I so choose?
Or do I have to use proprietary hubs and software to control the devices?
In short, are there any end user hostile features or can I use the devices like how Zigbee works?
I've done everything using old Particle Xenon boards (nrf52840 microcontroller) and didn't encounter anything 'user hostile'. Originally I had all the 'hub' stuff on one of the boards too, but the newer recommended way is to have the board be the 'radio' and to use a normal computer for the routing.
The matter network is just an IPv6 network, so I run coap server on my matter devices and then control them with command like 'coap-client -m post coap://[<ipv6 addr>]/open_button'.
Yes but since it routes IPv6 and hubs are usually connected to the Internet when set up by the average consumer, it is very easy for Thread devices to "accidentally" gain Internet access.
My understanding is that it can never be accidental since all access is mesh-local by default. You'd have to install a border router capable of supporting NAT64 features — SmartFriends™, I think this generally not possible with consumer border routers, correct? — and then explicitly enable it for a device.
