The great firewall basically works in four different ways:

1. Completely blocks access to some platforms/IP (e.g., twitter, WhatsApp, Reddit, known TOR entry nodes, etc.)

2. Completely forbids some sort of traffic (e.g., VPN-like traffic)

3. Intercepts traffic to allowed platforms and applies keyword filtering

4. Huge moderation teams across state supported media platforms

Being able to hide traffic (e.g., special hidden TOR entry nodes whose traffic is designed to blend in with standard HTTP traffic) will allow users to bypass points 1 and 2. The physical transfer of storage as another user mentioned will also work to exfiltrate info.

Using euphemisms (Winnie the Pooh instead of Xi Jinping), non-text media (screenshots, photos and videos), will bypass point 3, at least until the moderator teams catches on.

Point 4 is trickier, but it's basically a race against the clock. Subversive content will be removed, but it might survive long enough to be saved and forwarded somewhere else.

None of this is secret... They are technical limitations (mostly), not really anything that can be done about it.

What I'm not sure is why point 4 is based on accept-by-default (i.e., they could block all publications until moderators reviewed it), but I guess it would put too much strains on the communication, instant messaging would become impossible, for example. So they have to live with it.