It's sad that premature regulation moved this into the sphere of politics and mandated-purchases (from companies with an incentive to lobby). That means any attempt to learn what's really going on, or make any incremental changes, is gummed up by the fog of adversarial, zero-sum politics.
Without the inflexible mandate, the rule could simply be: label whether your product has flame-retardants or not, and have evidence for your claims. Then the industry could diversify, some brands emphasizing flame-resistance and others lack of chemical additives. And, this diversity would mean we would gather more real-world experience about whether the flame-retardant units are really correlated with fewer/slower fires and fewer fatalities, or with more health/IQ/behavior problems in children and pets, or both, or neither.
But because of the uniform monoculture imposed by the premature politicization, we face a data white-out and policy is flying half-blind... with only indirect data on costs and benefits.
Maybe a way out would be to compel a bisection of the industry output: you must make half your units with, and half without, these additives, and clearly mark the units in each category. Five to ten years later, we'd likely have much better data about real effects.