gg5, prog

I think it's OK as an rK form.

@@ -26,0 +27 @@
+<gloss>without knowing it</gloss>

Based purely on references しらずしらず seems to be more common for 不知不識.
That said, 不知不識 itself is not common and I thus wasn't able to source any examples with confirmed readings (a shame that yourei.jp is gone).

不知不識（しらずしらず）： Nikkoku, Shinsen, Saito, Oukoku, Iwakoku, GendaiRK, Koj
不知不識（ふちふしき）： Daijs, Nikkoku

Google N-gram Corpus Counts
╭─ーーーー─┬─────┬───────╮
│ 不知不識 │ 371 │ 93.9% │
│ 不識不知 │  24 │  6.1% │add; [sK]
╰─ーーーー─┴─────┴───────╯
Nikkoku even has just these two for しらずしらず

Smk: もとは「｛不知｝｛不識｝」と書いて、返り読みをした。

.

Based on that I don't think 不知不識 should be dropped.
At the minimum it should be included as an [sK], though I don't see the harm in having it as an [rK] as it meets the threshold.

@@ -10,0 +11,8 @@
+</k_ele>
+<k_ele>
+<keb>不知不識</keb>
+<ke_inf>&rK;</ke_inf>
+</k_ele>
+<k_ele>
+<keb>不識不知</keb>
+<ke_inf>&sK;</ke_inf>

Since 不知不識/ふちふしき is a separate entry, with an xref to here, I think it can be dropped as a kanji form. It's at best a wierd gikun use of しらずしらず.

@@ -10,4 +9,0 @@
-<ke_inf>&rK;</ke_inf>
-</k_ele>
-<k_ele>
-<keb>不知不識</keb>

知らず知らず 345,792  97.6% 
知らず識らず   1,560   0.4% -rK (daijs)
不知不識　　     371   0.1% -rK (oukoku)
しらずしらず   6,605   1.9%

I think we can drop the [yoji].

@@ -9,0 +10 @@
+<ke_inf>&rK;</ke_inf>
@@ -12,0 +14 @@
+<ke_inf>&rK;</ke_inf>
@@ -20 +21,0 @@
-<misc>&yoji;</misc>

Adding yoji tags based on gakken, shinmeikai and-or jitenon
-*- via bulkupd.py -*-

@@ -19,0 +20 @@
+<misc>&yoji;</misc>