tip_income carrier support: return-count target -50.27% on certified N; selection-mass protection missing from launch config

Atomic residual decomposed from #451 (item 3). The tips dollar surface is solved; the carrier surface is not.

State on the certified N default (c3e378a-20260722T010408Z, receipts on #451)

• box_7_social_security_tips.amount@2024 fits at −0.01%; …return_count@2024 sits at −50.27% (within the loss cap, full gradient — a genuine support limit, not a cap artifact).
• Support: 549 carrier rows / 3.00M weighted vs the W-2 Table 4.B TY2020 anchor (6,038,613 returns) and Treasury TY2025 filing-season claims (>6M filers). The solve hits the dollar mass on roughly half the expected tipped returns ⇒ average tips per carrier ~2× reality; any per-worker tips analysis (e.g. no-tax-on-tips incidence by worker counts) inherits that ~2×.
• OBBBA tips probe: −$1.63B at 2026 law = 16% of the JCX-35-25 FY2026 score (−$10,121M) — the carrier gap, not the dollar mass, is now the binding constraint (#499 receipts).
• The planned tips selection-mass protection did not ship in N's launch config (keogh #446-pattern protection shipped for keogh only).

Fix ladder (atomic)

1. Selection-mass protection for tip_income in the next launch config (the #446 keogh pattern) so reselection cannot drop carrier mass further.
2. Donor-participation support widening: the 2026-07-22 forensics verdict attributes the carrier deficit to the sipp_tips donor participation concept (December-reference monthly tips ×12 undercounts annual any-tip participation); widen the donor participation concept or reselect support so the return-count target can bind without distorting per-carrier averages.
3. Re-measure both tips targets + the OBBBA probe on the rebuilt support.

Ties: #446 (carrier/mass protection class), #481/#487 (support-forensics family), #499 (anchors + probes, merged), #474 (wages-chain aging). Doctrine #492: this is support work, not a calibration knob.

Start by comparing the next launch configuration with the #446 keogh protection pattern, then inspect the sipp_tips donor participation concept identified in the 2026-07-22 forensics verdict. Done means reselection preserves carrier mass, the two tips targets are re-measured on rebuilt support, and the OBBBA probe is re-measured without distorting per-carrier averages.