Sorted from highest to lowest "priority":

• bug, all mods: fix itemHandler cap processing - when block/entity returns cap for direction-less item handler dont query direction-aware itemhandlers -> entire content will be duped
• possibly bug: check our classes extending AbstractArrow (and their callsites) if they correctly use AbstractArrow#pickupItemStack (they don't or they mimic it) so bow logic etc. work reliably
• bug, DO floating carpets have wrong sounds/should follow vanilla carpets
• block/item/entity types should switch to deferred registry (or anything else) to unify and clear the mess around said registries - ideally we should use macro methods when applicable to make porting easier
• rework player interaction permission handler 03b118d#diff-e3c763fa2a25fea1952fa9b7ae9985c9024fb03710b0c64274807e94759f4766R377-R399
• CompostRecipe -> data maps - compostables
• RFC: CustomRecipe -> normal recipe with normal serializers etc. with condition FALSE (so it never works for players), requires datapack breaking change
• unify sending registry_value.id when sending over network (ported many to resloc cuz it was easier that way, but we should be able to send numerical ids instead)
• all mods: ItemBlockRenderTypes#setRenderLayer
• stop creating/passing around FriendlyByteBuf - either use nbt or just do (de)serialization in off-thread from/toBytes

CompostRecipe -> data maps - compostables

I thought about that during the 1.20.4 port, but concluded that it wasn't sufficiently compelling because the existing system allows for automatic addition of items based on food/plant properties, which a data map system would not.

RFC: CustomRecipe -> normal recipe with normal serializers etc. with condition FALSE (so it never works for players), requires datapack breaking change

There's actually no requirement that vanilla recipes be usable by players, so you don't have to do anything special to disable that. There's no particular technical barrier to putting colony recipes into the regular recipe manager, but not really much benefit either. We're already handling synching and we would still require custom handling for loot tables, so it doesn't even let us remove much code.