Venezuelan heavy-sour exports return above 1M bpd → USGC light-heavy differentials widen → coking-capable refiners buy feedstock cheaper against a benchmark-priced product slate → Valero and Phillips 66 capture the spread → but Marathon explicitly opts out

The feedstock-differential leg of the refining trade, distinct from the crack-spread leg already filed as [[refining-bottleneck-to-refiner-crack-capture]] — Venezuelan heavy crude is back above 1M bpd of exports, heavy-sour discounts have widened, and Valero's Q2 2026 refining operating income tripled to $4.5B with management naming Venezuelan sourcing as a competitive advantage; Phillips 66 quantifies the sensitivity at ~$140M of annual EBITDA per $1 of WTI–WCS widening.

• **WTI–WCS narrowing back inside ~$8/bbl** while Venezuelan exports hold above 1M bpd — would mean the barrels are being absorbed without widening the discount, breaking step 2. *Machine-checkable: `wti_wcs_differential < 8`.*
• **Valero's Q3 2026 refining operating income falling back toward the $1–2B range** on stable throughput — would mean the Q2 result was crack-driven (already captured by the sibling chain) rather than feedstock-driven, breaking step 3/4.
• **Sanctions snap-back or a political reversal cutting Venezuelan exports below ~0.6M bpd** — removes the forcing function, though not the Canadian/OPEC+ drivers.
• **Maya/Mexican heavy recovering sharply while Venezuelan barrels arrive** would be *supportive*; the reverse — Canadian egress constraints easing heavy supply out of the USGC — would tighten the differential against the chain.
• **PSX abandoning or being denied direct PDVSA purchase authority** — would weaken the "sourcing alpha" distinction that separates VLO/PSX from MPC.
• 2026-08-14-autoresearch-usgc-coking-capacity-binding-ceiling-or-not — Kpler's $20/bbl vs $15/bbl sour coking margin with explicit Venezuelan attribution (the disinterested quantification that moved step 3 to `confirmed`); Argus on the 4% coke-production decline (answering the coking-ceiling question NO) and the ⚠ petcoke-oversupply offset; and the ⚠ note that western Canadian heavy recovery shares the differential attribution, which softens the sanctions-reversal falsifier.

• ⚠ **A supply withdrawal offsets the supply return, and no source nets them.** Mexican Maya imports have collapsed to **22,000 b/d, a six-year low** — Maya is the traditional Mars alternative for USGC coking refiners. Net heavy-sour availability on the Gulf Coast is therefore materially less improved than the Venezuelan headline implies. *This is the sharpest open weakness in the chain.*
• **The differential is over-determined.** Canadian growth, US Gulf offshore, the OPEC+ unwind and Venezuela all push the same direction. Attributing refiner margin to Venezuela specifically is not supported — which matters because a Venezuela-specific falsifier (sanctions snap-back) would remove only one of four drivers.
• **Durability of the ~1M bpd level is contested.** CFR argues that "increasing Venezuela's oil output will take several years — and billions of dollars." The recovery may be a plateau, not a ramp.
• **All differential levels in this chain are Q1/Q2-dated.** No August 2026 WCS/Mars/Maya print was found. The chain is currently marked off stale differentials.
• **The adversarial check is missing.** No analyst on the Valero call probed whether the discount is sustainable or already priced, and management "did not address conditions that would narrow heavy crude discounts."

• **VLO is the primary expression, PSX the secondary.** Valero is the largest US consumer of the barrel and says its run rate will exceed its historical maximum; PSX has gone one step further by seeking to bypass intermediaries and buy direct from PDVSA, which sharpens the economics if it clears.
• **PSX supplies the dollarization the chain needs**: ~**$140M of annual EBITDA per $1** of WTI–WCS widening. That converts a differential print into a P&L estimate and gives risk a watchable number. ⚠ Sourced secondarily — verify against the PSX transcript before using as a signal threshold.
• **MPC should be held as differential beta, not as this chain.** It benefits from wider spreads (Hessling expects them to widen) but is not executing the Venezuelan leg. Treating VLO/PSX/MPC as one basket conflates two exposures. This is the correction the 2026-08-11 hypothesis sketch needed.
• **Spine-distinct from refining-bottleneck-to-refiner-crack-capture.** That chain is about product-side capacity (crack spreads at seasonal highs because refining capacity is short). This one is about input-side cost (feedstock discount widening). They can move independently, and they cross-read: a refiner short of capacity *and* long of cheap feedstock is doubly advantaged, which is roughly what Valero's quarter looks like.
• **Valero adds capability into the window**: the $230M FCC unit optimization at St. Charles starts up in **Q3 2026**.