Seems to be about what Uniscribe does.  Not exactly.  But close enough.
More consonants will start a new cluster.

A few scripts went way down in failures.  In particular:

  - Devanagari failures went down from 490 to 56.
  - Telugu went down from 113 to 49.

Other scripts went down slightly or didn't change.  New numbers:

BENGALI: 353908 out of 354285 tests passed. 377 failed (0.106412%)
DEVANAGARI: 693572 out of 693628 tests passed. 56 failed (0.00807349%)
GUJARATI: 366485 out of 366506 tests passed. 21 failed (0.00572978%)
GURMUKHI: 60750 out of 60809 tests passed. 59 failed (0.0970251%)
KANNADA: 950730 out of 951913 tests passed. 1183 failed (0.124276%)
KHMER: 298613 out of 299124 tests passed. 511 failed (0.170832%)
MALAYALAM: 1046881 out of 1048416 tests passed. 1535 failed (0.146411%)
ORIYA: 42320 out of 42329 tests passed. 9 failed (0.021262%)
SINHALA: 271333 out of 271847 tests passed. 514 failed (0.189077%)
TAMIL: 1091837 out of 1091837 tests passed. 0 failed (0%)
TELUGU: 970524 out of 970573 tests passed. 49 failed (0.00504856%)

Some of the remaining Telugu and Devanagari issues seem to be Uniscribe
eating Anusvara when placed before a non-joiner.  Ouch!

Malayalam failures go way down.  Other scripts benefitted slightly too.
Sinhala had one or two test regressions, but...

That's really what Uniscribe does, and explains a lot of pecularities of
Halant,ZWNJ before the base.

Sent Telugu from 1% failures to 0.03%.  Improved Kannada and Malayalam
slightly.  Fixed half of Bengali, and did NOT break anything!

In Sinhala, Rakar is formed by Al-Lakuna,ZWJ,Ra.  If you put that at the
end of a Consonant,Matra syllable, you get a dotted-circle from
Uniscribe.  Apparently adding a ZWJ before the Al-Lakuna "fixes" that.
And people have been encoding that sequence...  So, allow a forced
"ZWJ,Virama,ZWJ,Ra" sequence at the of syllables.

Fixes some 100 or more of Sinhala failures.  Now at 622 only (0.23%).

The sequence <ZWJ,Al-Lakuna,ZWJ> is used in Sinhala to explicitly ask
for Rakar.  Fixes two-thousand Sinhala tests.  Not many left.

Really this time.

Fixes another 18 Khmer tests.

Fixes 1 more Devanagari test!

Fixes 2 out of 560 Devanagari failures.  AND:
Fixes 1 out of 2 Tamil failures.

Refactoring Indic machin.  No semantic change.

It's a visual Repha.

Still not positioning logical Repha as occurs in Malayalam.

Another 200 Khmer failures fixed.  547 to go.  That's better than
Devanagari!

Down another 500 or so Khmer failures!

Except that there may be a ZWNJ before a Register Shifter.

Amend the syllable structure to allow a final subscripted consonant
(Coeng+C) and a final subscripted independent vowel (Coeng+V).
Fixes another 2k of Khmer failures.

No characters recategorized yet.  No semantic change.

Fixes another 6% of the Khmer failures.

So we can use them to switch default visibility to internal if desired,
and use these to make only declared symbols public.

For dotted-circle independent clusters, Uniscribe does no Reph shaping
for the exact sequence Ra+Halant+25CC.  Which also is the only possible
sequence with 25CC at the end.

Uniscribe allows multiple Halant+Consonant after a Vowel.
Tests:
↦       * U+0905,U+094D,U+092B,U+094D,930,94d,930

Uniscribe allows up to two nuktas per consonant and one per matra. It does so
indepent of whether the consonant already has a nukta in it.  Tests:

        * U+0916,U+093C,U+0941
        * U+0959,U+093C,U+0941
        * U+0916,U+093C,U+093C,U+0941
        * U+0959,U+093C,U+093C,U+0941
        * U+0916,U+093C,U+093C,U+093C,U+0941
        * U+0959,U+093C,U+093C,U+093C,U+0941
        * 915,93c,93c,,94d,U+0916,U+093C,U+093C,U+093e,93c,93c

See comments.

This is to incorporate it into GSUB/GPOS processing.

As reported by datao zhang on the mailing list.

https://www.microsoft.com/typography/otfntdev/devanot/shaping.aspx

Makes number of failures against Uniscribe with hi_IN dictionary from
OO.o to go down from 6334 to 4290.  Not bad for a one-line change!

Mozilla Bug 729626 - ASAN: heap-buffer-overflow HTML

Find the base consonant and apply basic Indic features accordingly.
Nothing complete, but does something for now.  Specifically:
no Ra handling right now, and no ZWJ/ZWNJ.

Number of failing shape-complex tests goes from 174 down to 125.

Next: reorder matras.

Still not much there.

Using an incredible tool called Ragel.