ECM I'm not quoting your post, it's too long, but you have multiple logical fallacies contained therein. First, you assume spending more spell slots on it is desirable. Spending fewer spell slots on it is more spell slots available for other things.
My comments are not based on spell slot cost, but they are relevant. One argument presented is that you use a 1st level slot to cause a higher level effect, but that is only true when the enemy fails. If the enemy saves you are using more spell slots to cause no effect.
If the enemy has an 80% chance to make the save (needs to roll a 5), it will cost on average 6 levels worth of slots to reverese the outcome one time (4 first level slots and one second level when you are out of 1st level slots). Further it costs 5 reactions. So if you are using it with banashment, on average it will cost 6 levels worth of spell slots to land one mre 4th level banishment using silvery barbs.
Now if the enemy has a 20% chance to save as in your example, then yes it is very powerful to use statistically, but you suffer from opportunity.
Think of this like feather fall. Feather Fall is the most powerful spell available if your party is plummeting from 500 feet altitude. It will flat stop a certain TPK. Nothing, not even wish, is better in this circumstance (wish is not better because it is not a reaction). So when that happens feather fall is THE BEST spell you can have prepared. If that does not happen though then having feather fall is not useful, or is of limited use (like say you are falling 20 feet). Silvery Barbs is the same. Having it available when an enemy who is about to kill your party rolls a needed 19 on a save or suck is HUGE, but that does not happen often and if it does not happen then it is not relevant to have it.
Second, you ignore that you do not need to choose to use it until and unless someone has made the save.
This makes it weaker agaisnt foes with good saves. You can only use it when the enemy has already made a save, which means the slot cost and action cost is high. Perhaps not as high as the original spell cast (and wasted), but it will likely be a failure if the enemy has a good saving throw. Further because this is after the first roll suceeded the enemy is MORE likely to save than if he just had flat disadvantage.
For example - the enemy needs a 7 to save (actual from red Dragon-banishment example). IF you give him disadvantage on the save his chance of failing is 51%, meaing he will fail more than he succeeds. Howver if he already suceeded once then his chance is only 30%. Meaning when you make the choice to use it he will most likely succeed.
So if there is an enemy with a poor save who makes the save in question, it will not "usually fail", it will usually succeed, as the attempt at using it for that begins in the state at which the enemy has already made the save in question.
No it is extremely powerful in that case, but that is rare and being rare makes it SITUATIONAL .... like the Feather Fall example above.
You don't have a lot of high level spell slots, if the enemy has a poor save most of the time having this spell prepared will be irrelevant. Not all the time, but most of the time.
Third, you ignore the impact of a successful spell is often a combatant being entirely eliminated. The most power SB ever is when the DC is a natural 20, because then you took someone you -know- made the save, got lucky, and remained in the fight to potentially cause huge problems, and you were able to make it exceedingly likely that, due to your use of a 1st level spell, they are instead eliminated from the fight.
It takes 35 full adventuring days to reach 20th level.
About half of those days are 11th level or above.
This means going from level 1 to level 20 you will cast approximately 60 spells of 6th level or higher (assuming you use all your spell slots).
If every single one of them targets a creature that needs a 20 to save, on average this will happen THREE TIMES with a high level spell going from level 1 to level 20.
This assumes you use ALL your spells to target someone that needs a natural 20 to save, AND it assumes every single high level spell you cast is a save or suck. No summoning, no Wishes, no simulacrum ....
Three times!
Finally, you act as if you can only use it on spells you yourself are casting. It's not that 'most days you won't use it because no enemies are failing your saves anyway' (which sound like pretty easy days to me and a desirable outcome!), it would have to be that 'no enemies are failing -any- saves from -any- abilities of your party members'.ty
This is true, but not enough to make it OP. I played in a gorup that went to tier 3 with 3 casters that had it.
Also worth noting that Legendary Resistance makes this MORE effective, since normally to get a chance at burning one you need to connect with a potent enough effect to be worth the burn, and you want to burn through those ASAP. This gives you another chance at burning one with extreme action efficiency, helping guarantee you peel those away ASAP so your spells can actually land. Not saying spending your reaction on it is a 'free' thing or without risk, but your logic and calculations are flawed.
You are trading your reaction for a CHANCE to burn a legendary and if he succeeded on the original save, it is probably not a good chance.
Do that against a Lich and you can't counterspell, do that against a dragon and you can't cast absorb elements.
I would like an exampel of an enemy and a spell where you think this would be a good idea for say a 14th level caster.
Also keep in mind you only have 4 1st level slots AND you need to be within 60 feet to cast it.