Bugzilla

Summing up the situation here -- the map on the live site here happens to use `zoom`, `polygon`, and `viewBox`, which means it's susceptible to two bugs, which happened to cancel each other out, sort of. And we fixed one, which is why the issue is now visible.

Specifically:
(A) we have a bug where we *fail to apply `zoom` to `polygon`.  (This goes back to bug 1854441 as noted in comment 7 and I think it's the defect that we should focus on here.)
(B) until recently we had a bug where we'd **redundantly apply the `zoom`** to the SVG-viewport-as-a-whole (or something along those lines) if a `viewBox` attribute is present.  That issue was bug 1905023.

On this particular site, issue (A) was essentially getting papered over by issue (B).  When we fixed issue (B) (in bug 1905023), issue (A) became visible.

Summing up the situation here -- the map on the live site here happens to use `zoom`, `polygon`, and `viewBox`, which means it's susceptible to two bugs, which happened to cancel each other out, sort of. And we fixed one, which is why the issue is now visible and why it manifests as a regression from bug 1905023 (per comment 1) even though really it's not indicative of a problem in that bug's patch (I think).

Specifically:
(A) we have a bug where we *fail to apply `zoom` to `polygon`.  (This goes back to bug 1854441 as noted in comment 7 and I think it's the defect that we should focus on here.)
(B) until recently we had a bug where we'd **redundantly apply the `zoom`** to the SVG-viewport-as-a-whole (or something along those lines) if a `viewBox` attribute is present.  That issue was bug 1905023.

On this particular site, issue (A) was essentially getting papered over by issue (B).  When we fixed issue (B) (in bug 1905023), issue (A) became visible.

Summing up the situation here -- the map on the live site here happens to use `zoom`, `polygon`, and `viewBox`, which means it's susceptible to two bugs, which happened to cancel each other out, sort of. And we fixed one, which is why the issue is now visible and why it manifests as a regression from bug 1905023 (per comment 1) even though really it's not indicative of a problem in that bug's patch (I think).

Specifically:
(A) we have a bug where we *fail to apply `zoom` to `polygon`.  (This bug persists in current Nightly, and it's what testcases 2 and 3 reveal; and it goes back to our initial `zoom` impl in bug 1854441 as noted in comment 7, and I think it's the defect that we should focus on here.)
(B) until recently we had a bug where we'd **redundantly apply the `zoom`** to the SVG-viewport-as-a-whole (or something along those lines) if a `viewBox` attribute is present.  That issue was bug 1905023.

On this particular site, issue (A) was essentially getting papered over by issue (B).  When we fixed issue (B) (in bug 1905023), issue (A) became visible.