PF2E - Here is an excerpt of the ability:
Captivating Song
Tags: Auditory, Concentrate, Enchantment, Incapacitation, Mental, Primal
Text:
The harpy cries out an eerie, compelling melody. Each non-harpy creature within a 300-foot aura must attempt a Will save to avoid becoming captivated by the harpy's song. The effect lasts for 1 round, but if the harpy uses this ability again on subsequent rounds, it extends the duration by 1 round for all affected creatures. Once a creature succeeds at any save against Captivating Song, that creature is temporarily immune to Captivating Songs for 24 hours.
Success: The creature is unaffected
Failure: The creature is 'Fascinated', and it must spend each of its actions to move closer to the harpy as expediently as possible, while avoiding obvious dangers. If a captivated creature is adjacent to the harpy, it stays still and doesn't act. If attacked by the harpy, the creature is freed from captivation at the end of the harpy's turn.
Critical Failure: As failure, but if attacked by the harpy, the creature can attempt a new save at the start of its next turn, rather than being freed at the end of the harpy's turn.
The rule is poorly written because interpretation sways its power considerably and is incoherent with Paizo's recommendations on monster ability design.
Rule Option 1 - the worst option:
Fascinated and Captivation are two separate effects. Even though another creature may break fascination with an action, the "Rider effect" of captivation persists, meaning the pc/npc is still crowd controlled by the harpy. As long as that harpy does not attack, and continues to sustain its song on each new turn, the effect will continue to persist indefinitely.
This is a problem, as it spells an immediate party wipe if all 4 pcs (assuming standard group size based on paizo's own recommendation) fail their will saves, and there are 2+ harpies present. Harpy 1 will sing indefinitely, harpy 2 will attack each player until they are dead, fascination may break on any creature action, but captivation does not. No saves, no counter play, you're facing a TPK.
Outside of balance, this also makes a mess of Paizo's own "ruling by inference" design. "Captivated" is not a status, a tag, or a effect. It's not defined. This is contrary to Paizo's recommendation that when designing monsters, their abilities should be tied to game mechanics to create fair, consistent rulings, throughout the campaign.
This ruling basically creates a status out of thin air, gives no details on what being "captivated" really means, defines no saves that can be made each turn, and no interactions with that status.
Additionally, the design of the ability makes being 'fascinated' purposeless. Fascinated may end, but the PC is still controlled, which is the intention of the spell.
Even in the event Harpy 1 attacks you, your character is still captivated until either the end of the harpy's turn, or in event of a critical failure, the pc can attempt to make a save at the beginning of their turn. This means the harpy can spend one action to sustain the song on multiple targets, an action to attack, and a third to sing again in an attempt to re-captivate the target they just hit.
Granted, you'd never attack with harpy 1 when you've got the party controlled, you'd just clean up with the adjacent harpies because the way the rule is written, nothing but the song's originating harpy can end the effect by attacking.
Even if this tactic is somewhat successful, it can still be used to devastating effect if it hits just 2 PCs in a group of 4. A Harpy that has captivated 2 out of 4 people can simple fly to the song's 300 foot distance (three hundred, you read that correctly), and avoid combat/concentration breaks by being out of range of any spells/abilities the party may try to use, and continually keep those party members crowd controlled while the remaining harpies mop up the 1-2 free PCs.
Absolutely busted.
Rule Option 2 - somewhat better:
Captivated relies on fascination. If fascination breaks, captivation ends.
This tactic is still powerful because it allows you to alter initiative and setup coordinated strikes on a party of PCs you captivate. As long as you don't use hostile actions, you can use multiple harpies to fish for a captivation (via attacking, breaking fascination, and attempting to reapply it with a new song), and if landed on all players, realign the initiative order to have 1 harpy at the top, all PCs in the middle, all other harpy's at the end. Using this strategy you reposition all pcs into a single area, delay all non-singing harpies to the end of the initiative ensuring PCs do not get to act, then descend upon one PC at a time with your harpies to kill them. Harpies are not brilliant, but they're smart enough to act like a pack of lions when hunting pray. In their description within the monster rulebook they enjoy attacking prey, or watching attacked prey who are captivated.
At the top of the initiative, attempt to re-captivate the players again and repeat the process.
This ruling is more consistent with Paizo's own recommendation of tying monster abilities to tags and ruling by inference, rather than creating status effects out of thing air with no rule support behind them. However, it still sucks because there is a large amount of abuse that can occur via tactics to spike this creature's power well over its intended monster design. Yet at least with this ruling, there is at a minimum some counter play that we can achieve to make this monster more balanced.
In my opinion, it's time to revisit fascination and make it do something useful if the power of this status is not strong enough to design monsters around. This is preferable to adding on "riders" that are not well defined, and are poorly balanced.
The harpy is the epitome example of relying on save or suck mechanics, or having to counter them with 'hero points'. This design is not enjoyable and very gimmicky.
