It appears that the natural generation of snow layers in Extreme Hills biomes and such uses a 32-bit float in its generation somewhere, causing it to break somewhat when far from the origin. This can be seen in vanilla by teleporting over 16777216 blocks out, preferably on both axes, and finding an Mountains biome (or just generating a custom world with exclusively Mountains) - any areas that would have had patchy areas of snow in vanilla will now be slightly more angular than they would normally be, being composed of more straight lines and squares.
This first started happening in 16w02a - versions prior (13w36a to 15w51b) are completely unaffected by this. For example, here's a screenshot of 1.8.2-pre4 (left) and 1.15.2 (right), with seeds provided:
[media][media]
I am not sure if this bug can cause crashes like MC-185925 did.
Related issues
relates to
Attachments
Comments


There are some obvious differences at the same coordinates on the same seed on a different version. Trying to narrow down where this first happened but it's somewhere between 1.8.2-pre4 and 1.15.2.
[media][media]

Can reproduce in 1.16.2 Pre-1

After doing a bit of experimentation in modded 1.16.1 it indeed appears to be the rain/snow or temperature gradient itself that loses precision.

Some more experimentation screenshots on a modded instance:
[media][media]
[media]
[media]
[media]
[media]
[media]
[media]

Can confirm in 1.17.1.

Replicated in 1.19.2 but with beach biome spawning in the middle of an birch biome with no oceans near it

That's not what this is about...

Can confirm in 23w03a

Can confirm in 23w03a

Can confirm in 23w05a

still in 1.19.4 and 23w13a

Can confirm in 1.20.1, 1.20.2 and 23w42a.

Can confirm in 1.20.4 and 24w05a and 24w05b.

Confirmed in 1.21.3

Can confirm in 24w45a

Mojang, please add a gamerule to remove the world border. Literally nothing is stopping you from doing that, as you can see in this reference video.

Confirmed in 1.21.4 prerelease 1.

Confirmed in 1.21.4 prerelease 2

Can confirm in 1.21.4 release candidate 3.

Confirmed in 1.21.4