The iridium gearbox recipe works at 50% the speed of the basic gearbox recipe. I feel that the output should be balanced, as is the case for eg. nitinol mesh or electronic components.

Looking around I also see that the aluminium-ceramic recipe for low density structure also produces at half the speed of the other 2 ground based recipes.

There is a vast difference in the production rate of drive belts - I assume that this one is deliberate, as it is nearly an order of magnitude? For my arbitrary assembler, 2.5/s compared to 20/s

Iridium heat shield is produced at twice the speed of the other two recipes for heat shield.

I think these recipes should produce output at the same rate. It appears that for the rest of those I checked (ie. I have unlocked) there has been a deliberate effort to eg. double the crafting time for recipes which produce 2 items instead of 1, which makes me think that the imbalance in those I listed is accidental. Except maybe the drive belts - but I'd like to draw attention to that anyway, in case it is a mistake.

Iridium heat shield is produced at twice the speed of the other two recipes for heat shield.

That's straight from SE.

Looking around I also see that the aluminium-ceramic recipe for low density structure also produces at half the speed of the other 2 ground based recipes.

I was already looking into this one (well, LDSs in general) based on another bug report; what happened here seems to be that 248k copies the vanilla LDS recipe at some point, then SE cuts the time in half. I'm planning to suggest that 248k change the time to match when SE is enabled.

There is a vast difference in the production rate of drive belts - I assume that this one is deliberate, as it is nearly an order of magnitude? For my arbitrary assembler, 2.5/s compared to 20/s

This one, I think was probably intentional, as a balance against the higher complexity of making the materials.

The iridium gearbox recipe works at 50% the speed of the basic gearbox recipe. I feel that the output should be balanced, as is the case for eg. nitinol mesh or electronic components.

That does seem to stand out; the clear trend in iridium recipes is the same output speed but lower material costs. And a nominal upgrade being slower is unusual.