I think we need to do some work on the thresholds, but I agree that in this case <3% is a bit low.

@@ -10,0 +11 @@
+<ke_inf>&sK;</ke_inf>

I didn't have (standard) okurigana variants in mind when we came up with that threshold. Since they're recorded in the kokugos (the larger ones, at least), I've always considered them to be in a different category from 交ぜ書き forms, forms containing katakana instead of hiragana, etc. Interested to hear what others think.
On 1614070, it was tentatively agreed to use a 3% threshold (my proposal) for standard okurigana variants but this was a year before the discussion on 2839852 where we decided on the 20% threshold.
Maybe 3% is too low but I feel that 20% is too high for these forms.

"か･わる" is the okurigana form favored by the joyo list.

I thought we were doing <20% for variants that only differ due to kana usage. I don't see a lot of value in keeping something like 風変り visible, but I'll keep that threshold in mind if that's what everyone else wants to do.

gg5, prog, luminous

I think the sK threshold we're using for okurigana variants that are supported by the kokugos is <3%. 風変り is in daijr/s.

@@ -11 +10,0 @@
-<ke_inf>&sK;</ke_inf>
@@ -29,0 +29,2 @@
+<gloss>odd</gloss>
+<gloss>unusual</gloss>
@@ -30,0 +32 @@
+<gloss>peculiar</gloss>

〈風／ふう／フウ〉〈変（（わ）り）／がわり／ガワリ〉

Google N-gram Corpus Counts
╭─ーーーーー─┬─────────┬───────╮
│ 風変わり　 │ 163,031 │ 94.2% │
│ 風変り　　 │   6,952 │  4.0% │ - sK
│ 風変　　　 │   1,781 │  1.0% │
│ 風がわり　 │     660 │  0.4% │ - sK
│ ふう変わり │      59 │  0.0% │ - add, sK (seen in the sankoku definition for 奇抜)
│ ふうがわり │     672 │  0.4% │
╰─ーーーーー─┴─────────┴───────╯

@@ -10,0 +11 @@
+<ke_inf>&sK;</ke_inf>
@@ -13,0 +15,5 @@
+<ke_inf>&sK;</ke_inf>
+</k_ele>
+<k_ele>
+<keb>ふう変わり</keb>
+<ke_inf>&sK;</ke_inf>