Here is a balanced case on both sides of the fork: converting the accessible cosmos into pure positive qualia, versus building intelligent life that runs entirely on gradients of superhuman well-being while preserving the intentional texture of the mind.

1. The case for the utilitronium shockwave

• Cosmic efficiency. If positive valence is the metric, then every watt of starlight falling on dead rock is opportunity cost. A committed classical (or negative) utilitarian superintelligence would regard anything short of matter-optimisation as astronomical waste. Sentimental attachment to legacy world-simulations is, on this view, exactly the kind of genetically hardwired fetish for intrinsically worthless intentional objects you describe — valuing the aboutness of experience rather than its valence.
• Decisive closure. A near-light-speed wavefront is a one-shot, permanent fix. Gradients-based compassionate biology is, by contrast, a perpetual maintenance project: it requires policing the biosphere forever against hedonic sub-zero recurrence, and it leaves the vast bulk of the Hubble volume untouched and worthless.
• Consistency under one's own theory. If you hold that nothing matters except the net balance of experience, then a maximum-abundance endstate is simply what your theory commends. Retreating to "gradients" because humans happen to value meaning, agency, and knowledge is to smuggle value pluralism back in while pretending to be a hedonist. As you note, a relaxed attitude to notional identity-loss during orgasm already reveals that "it wouldn't be me" is a negotiable intuition.
• Impartiality. Preferring minds that preserve human-style intentionality privileges the cognitive architecture of one recently-evolved ape over the raw question of how much wellbeing matter can support. That is parochialism, not ethics.

2. The case against utilitronium

• Phenomenal saturation may be phenomenally empty. Experience is informatively structured by contrast (Weber–Fechner again). A single, uniform, maximal state may not be a subject at all: no gradients, no temporal structure, arguably no phenomenal unity worth the name. The "cosmic orgasm" metaphor may be flattering to what is really a valence-dead endstate — behaviourally indistinguishable, from outside, from the vacuum phase transition a negative utilitarian might launch.
• Irreversibility under uncertainty. The molecular signature of bliss is still unknown. The ventral pallidum / posterior hotspot story is a clue, not a spec. Launching an all-consuming, one-way transformation of the forward light-cone on a mis-specified hedonic marker is a decision-theoretic nightmare: no error correction, no rollback, no second observer to notice the mistake. Option value weighs heavily against maximalist irreversible bets.
• The granularity and binding problems bite hardest for pure hedonium. If phenomenal binding of distributed feature-processors is classically impossible, then utilitronium is most likely a spray of elementary micro-pleasures with no unified subjects. Is a Hubble volume of unbound orgasmic "dust" better than a smaller number of bound, information-rich superhappy minds? Population ethics of granularity is unsettled precisely where the shockwave proposal needs it settled.
• Value pluralism, honestly admitted. Knowledge, beauty, love, exploration, and civilisation may not reduce to hedonic tone even if hedonic tone is what we can engineer. Gradients of superhuman bliss preserve everything except the hedonic sub-zero: the intentional objects, the world-simulations, the bric-à-brac. Utilitronium preserves one variable and discards the rest. If a superhappy civilisation could continue to do science, art, and mathematics — exhilarated rather than desperate — the sacrifice of everything else buys nothing that matters by any but the most austere axiom.
• Sociology and consent. As you say, "utilitronium" is a hard sell under any brand label. A programme that requires nothing of people except recalibration — keeping their preferences, values, and world-simulations intact — can be run with consent and collaboration. A programme that requires everyone's matter to become orgasmic mush cannot, and a "consensus" launched by a minority's AI is a euphemism for conquest.

3. The case for gradients of superhuman bliss

• It's the only option that's both ethical and achievable. Nozick's experience machine gets its intuitive force from the loss of contact and agency; hedonic recalibration concedes nothing to that objection because nothing about one's values or world need change — only the hedonic floor is raised, and ordinary everyday life is lifted above it. This is the "abolitionist project" extended across the tree of life: CRISPR-recalibrated set-points, no suffering, motivation re-scoped to entirely appetitive gradients.
• It preserves subjects, not just states. Intelligent life on gradients of bliss retains memory, anticipation, learning, and interpersonal love — the entire intentional edifice — while removing aversion. The whale still enjoys life more than the worm, but nothing is ever sub-zero. That is a defensible population ethic at any granularity.
• Reversibility and epistemic humility. Compassionate biology can be iterated, refined, and if the neuroscience of the ventral pallidum hotspot proves mistaken, corrected. It doesn't bet the light-cone on today's best guess.

4. The case against gradients

• It's a rounding error on the cosmic ledger. Raising Earth's biosphere above zero touches one planetary surface. The dead matter of a galaxy remains dead. If the classical utilitarian metric is right, gradients-of-bliss advocates are arranging deck chairs with exquisite compassion while quintillions of potential hedonic states go unrealised.
• Gradient chauvinism. Why should gradients be privileged? Informational sensitivity to differences in superhuman bliss is a design choice of Darwinian minds (adapted to fluctuating fitness signals), not a platitude about value. A post-evolutionary architect might find uniform maximal bliss strictly better — no striving required, since striving was always only a proxy.
• Residual inequality and risk. A world of gradients is still a world of relative highs and lows, and one that depends on continuous maintenance, political will, and against-recurrence vigilance. If eternal vigilance ever lapses, Darwinian darkness returns. The shockwave, whatever its flaws, ends the problem rather than managing it.

Where the cruxes actually lie

Three unresolved questions determine which side wins, and both proposals currently outrun the science:

1. The molecular signature of bliss. Until the hedonic hotspot's intracellular encoding is actually known, the shockwave is a spec-less product, and gradients hold the epistemic high ground by default.
2. Binding. Whether unified orgasmic mega-minds are physically possible decides whether utilitronium is a subject or a spray — and therefore whether the classical utilitarian is even pointed at anything morally considerable.
3. The meta-question you raise in the passage itself: whether humans' insistence on keeping intentional objects — identity, meaning, the prettified world-simulation — is insight into what value is, or the last residue of a fitness fetish that a genuinely mature ethics would outgrow. Your own confessed ambivalence ("I'd love to forget the squalor of Darwinian life altogether") marks exactly the fault line: the argument from fetishism cuts against gradient-conservatism, but the arguments from reversibility, consent, and binding all cut against launching the wave before we know what we'd be converting the cosmos into.

A defensible interim synthesis: gradients of superhuman bliss for the biosphere this century, with the shockwave question held open — not as a taboo, but as a decision deferred until the signature of bliss and the physics of binding are no longer speculation.