Why does "SMITH VD V, JOHN" lose the V from the family name when "SMITH V, JOHN" keeps it? #1080

rules/P9 (a particle that is also a suffix word, beside a family comma) says a lone letter behind such a particle reads as it does in the same part without the particle: a capital is a name word, so SMITH VD V, JOHN should read as SMITH V, JOHN does. Since 2.7.0 it does not. The particle (vd, van, van der) is left heading nothing, and the V becomes the suffix.

Measured on the 2.6.0 and 2.7.0 wheels and on master (25e698a plus #1079). The comma-less John van V reads family van, suffix V in both 2.6.0 and master, so it is not part of this.

Where it came from. Bisected over the 32 first-parent commits in v2.6.0..v2.7.0: the reading moves at 1f0b07c, "Ask one predicate for every roman-numeral rule (#752, phase 2f) (#886)". Its parent, 2af8da8, reads family SMITH VD V. The trailing numeral's words-to-spare count appears to count the unchained particle as a name word to spare, so the V is read as the suffix. No decisions entry (P9, S10, C3, P2) records or accepts the move. Nothing caught it because P9's example for this sat in prose rather than in a fence. #1079 adds it as a fenced example with a deviates: marker naming this issue.

Proposed rule shape (P9, unchanged):

• When: a word of both the particle and the suffix vocabulary stands before a family comma, with a lone capital letter behind it.
• Then: the letter reads as it does in the same part without the particle: a capital is a name word, and the particle heads it.
• Because: before a family comma the comma has already named the family; a particle with a name word behind it is a surname particle there.
• Except: with a suffix word or nothing behind it, the particle reads as the suffix (SMITH VD, JOHN → suffix VD), as P9 already says.

Fix: in 3.x. The numeral predicate's count should not count a particle that has a name word behind it before a family comma as a word to spare. When the fix lands, remove the deviates: marker from P9's example in the same PR.