I proposed the initiative for spells to be d6, d8, d10 based on the number of components to my players as well. I also suggested that bonus actions will not contribute to the initiative calculation. This includes casting spells as a bonus action. I also suggested that a spell casting initiative could be based on 1d4 or 1d6 per component used. Finally, I suggested that if a character cannot take an action when their turn comes, then rolling a d6 and adding it should allow them to take their turn later.
I then looked up a few spells to get an idea of which ones use all three and which ones use two. I feel a caster will know which spell they will want to use in advance enough to shift the flow of combat. There are several spells that use only two components. There are several spells that use all three. A player may look up the number of spell components on a few spells to determine if they are going to cast 2 component spell or a 3 component spell. If they roll for a 2 component spell and then find they need a 3 component spell, then when their turn comes they can roll a d6 and shift their turn to later in the round.
I am concerned that giving a player a chance to change their mind will slow the turn down. I gave the caveat that if turn time is dragging on due to indecision then I'll remove the optional d6 to shift turns later into the round.
You seem to contradict yourself a bit with these statements:
I feel a caster will know which spell they will want to use in advance enough to shift the flow of combat.
[...]
I am concerned that giving a player a chance to change their mind will slow the turn down.
I'm not talking about a player "changing their mind" - I mean when their chosen option simply isn't viable at the point their turn comes up. Not declaring the specific spell ahead of time, even though it's likely they know what they're going to cast, allows for them to switch to a backup spell, adapting to the changing circumstances of the battle.
Making them stick to what they've declared in the initiative phase could also have the effect of creating indecision - they've got to cast the spell they choose, so what's the best spell to pick if things change during the round?
Would you also make archers declare a target in advance, and if that target dies or becomes no longer visible, make them lose their action as well? A ranged attacker can currently simply pick the best/preferred target when their turn comes up. They may still end up with no targets, but that will happen far less than having to declare one. Same with casting spells.
And if you're worried about adding time for a player to figure out their turn, why, oh, why have a system where you have to stop and look up the specifics of the spell each time one is cast in combat? Think of the poor DM who has more than one spell caster NPC fighting the party.
The "adding a d6" to take a turn later ... nope. If they can't take the action now, it's still unlikely they'll be able to do it later. Unless you're suggesting they can also make a new declaration of what they're doing (potential for abuse there.) I mean:
DM: Any 9's?
Player 1: That's me, but the orc I was going to charm is dead.
DM: OK, roll me a d6 to act later.
Player 1: I got a 1.
DM: Alright. Any 10's?
Player 1: Uh, that's me, and the orc I was going to charm is still dead.
Like someone else said, this UA system is slightly more simulationist than the current initiative system. The question is, how far down the simulation rabbit hole to you want to go with further tweaks? Especially with non-real world things like casting spells.