I'd push back on the framing—this seems designed to make ethics binary when they're usually messier.

The honest tensions:

• Deadnaming or misgendering someone causes documented psychological harm. It's not costless just because it seems smaller than "1 million lives"
• But yes, most ethical frameworks do allow harming individuals when preventing vastly greater harms (trolley problem logic)
• The real question isn't whether the math works in principle, but whether this scenario is realistic

Where it gets complicated:

• In actual emergencies (fire evacuation, medical triage), people's pronouns genuinely don't matter—you're not choosing to misgender them, you're deprioritizing everything non-essential
• But a hypothetical where you must actively misgender someone to save millions is oddly specific. That suggests the misgendering itself is doing the lifesaving work, which is hard to construct realistically
• If misgendering were actually necessary, most trans people I've seen discuss ethics would likely agree it's justified—but they'd also note this gets weaponized to justify everyday misgendering that saves no one

My take:

The math works in principle (harm reduction can justify collateral damage). But I'd be skeptical of anyone reaching for this argument in non-hypothetical contexts, since real life rarely presents such clean trade-offs.

What's the actual scenario you're thinking through?