2.14.6

Someone's still seeing flicker at beaches, so I went looking in the shallow water instead of on the sand. 2.14.5 fixed the foam on dry land — that code doesn't even run where these new screenshots are showing it. Found two things there.

Foam was worked out once per block. The white surf ribbon at the waterline gets its strength from how close that spot is to land. That was measured once for the whole block and handed to all four corners, so every block wore one flat amount of foam with a hard step to the block next door. And the amount breathes with the wave — so each block brightened and dimmed as a whole square, out of step with its neighbours. Block-sized foam blinking along the shore. It's the same mistake as the wave heights in 2.14.3, one variable over. Now it's measured at the corners and blended across, so neighbours agree.

Blocks were dropping out of the sea entirely. There's a limit on how many water columns can be re-measured per frame — that's the fix that stopped the freezing near shores. When it ran out and a column had never been measured, it came back with "surface unknown". The sheet reads unknown as dry land, and dry land with no ground height gets skipped. So the block just wasn't drawn. A hole. And which columns land on the wrong side of that limit moves every single frame, so the holes blink. Near shores, because that's where the limit runs out.

Now it borrows the reading from a block next door instead. Costs nothing, and the sea stops having gaps in it.

Also: when a corner has nothing measured around it at all, it used to ask whichever cell was doing the asking. Two cells share that corner and each gave a different answer, so the surface tore there. Exactly the 2.14.3 bug, still alive on the one path I never looked at. It's a fixed value now, so both cells agree.

Being straight about this: I found these by reading code, not by watching the bug. Two of them are definitely wrong regardless and I'd have fixed them anyway. Whether they're the flicker someone's reporting, I don't know yet.

If it's still there on 2.14.6, the fastest thing is Visuals tab → Surface Cost Logging, then paste a few lines. There's a number in there for how many water columns got reused past their refresh. If it's not zero, that second bug is the one, and I'll know instead of guess.