本当	12361117	99.9%
本當	8357	0.1%

I agree with Marcus on this, and don't think this entry should be split at all; in fact the only change I'd make is to hide the 本當 form. Leaving it open for a bit, but I think the consensus is to stay with ほんと, etc. in this entry.

@@ -12 +12 @@
-<ke_inf>&oK;</ke_inf>
+<ke_inf>&sK;</ke_inf>

i do indeed think that multiple reading tags ([ck] (colloquial kana), [fk] (formal kana), etc.) would be helpful and if they're introduced the split would be unnecessary
if they're not, i still think i prefer preserving more info even at the cost of dispersing similar info

I understand the point you're making but I don't think the proposed changes would be a net plus to users, having two almost identical entries just needlessly spreads out the same info in my opinion. It's true that ほんと is [col] when written and like Robin said, having a [ck] reading tag might be helpful, if not necessary. You could also make the argument that we could make reading tags based on all or most of our already existing misc tags (e.g. [ksbk])  and that maybe we should be wary of "feature creep".

I don't think we need restrictions for 本當 as pronouncing the word ほんと apparently goes back to at least the Meiji period - Gakkoku for example has an 1887 example (with the kanji 真個, though!). 本當 was used in the Showa period up until the war at very least, see lab.ndl.go.jp
 (even if we didn't have evidence the pronunciation went back this far, the inclusion of 本當 is essentially a parenthesis so I don't think restrictions would have been necessary)

i wasn't referring to that
it's useful for the users to know that 本当/ほんと is [uk] and [col], as explained by the various kokugos
also, not splitting requires introducing restrictions as i don't think "本當" can be ever read "ほんと" which seem disliked when possible by the others (not me personally)
overall, i just think that splitting provides more value to the user

The kokugos don't have our 2/3 rule so that they have separate entries for 本当 vs ほんと here isn't really relevant.

A [ck] (colloquial kana) tag could be useful in cases like this.

sankoku:
〔話〕「ほんとう（本当）」の変化。
daijr:
「ほんとう（本当）」の転。会話語として使われる。
shinkoku:
〔「ほんとう」のくだけた言い方〕→ほんとう（本当）。

i think in this case it's best to follow the kokugos

It's just a tiny difference in prounciation. It's not really [col] when spoken like this in most cases. I know we're sidestepping the 2/3 rule more and more but I don't think this should be split out.

sankoku, iwakoku, gg5, daijr, meikyo, shinkoku, etc.

i think it's best to split ほんと/ホント, many kokugos seem to do so as well
i don't think 本當 can ever be read as ほんと
i've made ホントー/ホントウ search-only

@@ -21,9 +20,0 @@
-<reb>ほんと</reb>
-<re_inf>&ik;</re_inf>
-</r_ele>
-<r_ele>
-<reb>ホント</reb>
-<re_nokanji/>
-<re_inf>&ik;</re_inf>
-</r_ele>
-<r_ele>
@@ -31 +22 @@
-<re_nokanji/>
+<re_inf>&sk;</re_inf>
@@ -35 +26 @@
-<re_nokanji/>
+<re_inf>&sk;</re_inf>

daij, gg5, prog

Not an adverb.

@@ -37,0 +38 @@
+<pos>&n;</pos>
@@ -40,2 +40,0 @@
-<pos>&adv;</pos>
-<pos>&n;</pos>
@@ -50 +48,0 @@
-<pos>&adv;</pos>
@@ -59 +56,0 @@
-<pos>&adv;</pos>
@@ -61,0 +59 @@
+<gloss>real</gloss>

"ik" means irregular, which I think it is in this case. I think it should stay.

@@ -21,0 +22 @@
+<re_inf>&ik;</re_inf>

ほんと is in daijr as "会話語として使われる。", so just "col" at most, not ik.

@@ -22 +21,0 @@
-<re_inf>&ik;</re_inf>

ホントーに	31974
ホントに	6252892
ホントウに	19639
ホントー	39434
ホント	15625654
ホントウ	35218

Since we have 1611580 with the ホントー/ホントウ versions, I guess it should be consistent

@@ -27,0 +28,8 @@
+</r_ele>
+<r_ele>
+<reb>ホントー</reb>
+<re_nokanji/>
+</r_ele>
+<r_ele>
+<reb>ホントウ</reb>
+<re_nokanji/>

6 examples of ホント in Tanaka corpus
ngrams
 ホント	15625654
 ホントー	   39434
 ホントウ	   35218

@@ -21,0 +22,5 @@
+<re_inf>&ik;</re_inf>
+</r_ele>
+<r_ele>
+<reb>ホント</reb>
+<re_nokanji/>

It's officially a 形容動詞 (大辞林 has "（ 名 ・形動 ） ［文］ ナリ", but as is often the case, の is used more often than な. I don't think it hurts if we have both.

@@ -24,0 +25 @@
+<pos>&adj-no;</pos>
@@ -33,0 +35 @@
+<pos>&adj-no;</pos>
@@ -41,0 +44 @@
+<pos>&adj-no;</pos>

KM ngrams:
  本当の	238719
  本当な	15097

adj-no pos?

Daijr, GG5, etc. etc.

Glossing this effectively is a pain, as is the POS. Daijirin says "名 ・形動: and there are plenty of cases of "本当が.." so it's got to have noun in there somewhere.
Following GG5 and 新和英中辞典's sense split, but they overlap a bit.

@@ -26,0 +26,1 @@
+<pos>&adv;</pos>
@@ -27,2 +28,1 @@
-<gloss>honesty</gloss>
-<gloss>honest, truth</gloss>
+<gloss>truth</gloss>
@@ -30,0 +30,18 @@
+<gloss>actuality</gloss>
+<gloss>fact</gloss>
+</sense>
+<sense>
+<pos>&adj-na;</pos>
+<pos>&adv;</pos>
+<gloss>proper</gloss>
+<gloss>right</gloss>
+<gloss>correct</gloss>
+<gloss>official</gloss>
+</sense>
+<sense>
+<pos>&adj-na;</pos>
+<pos>&adv;</pos>
+<gloss>genuine</gloss>
+<gloss>authentic</gloss>
+<gloss>natural</gloss>
+<gloss>veritable</gloss>

@@ -27,1 +27,2 @@
-<gloss>truth</gloss>
+<gloss>honesty</gloss>
+<gloss>honest, truth</gloss>

http://en.wiktionary.org/wiki/本当

@@ -9,0 +9,4 @@
+</k_ele>
+<k_ele>
+<keb>本當</keb>
+<ke_inf>&oK;</ke_inf>