Submitted by pauliax, also found by BiasedMerc, NentoR, gjaldon, crypticdefense, zhaojohnson, twcctop, bigtone, b0g0, DanielArmstrong, fyamf, GoatedAudits, 0xCiphky, zigtur, xg, SBSecurity, lanrebayode77, blutorque, aslanbek, Aamir, araj, TheFabled, t0x1c, tapir, eeshenggoh, p0wd3r, peanuts, Greed, 0xordersol, 14si2o_Flint, guhu95, m_Rassska (1, 2), ustazz, maxim371, Fassi_Security, shui, mt030d, aman, rbserver, mussucal, josephdara, zzykxx, honey-k12, 0xnightfall, Maroutis, Aymen0909, OMEN, Stefanov, FastChecker, hunter_w3b, gesha17, baz1ka, kinda_very_good, carlitox477, 0xAadi, 0rpse, ak1, 0x73696d616f, 0xhacksmithh, ilchovski, LessDupes, adam-idarrha, siguint, 0xnev, 0xPwned, carrotsmuggler, KupiaSec, grearlake, and oakcobalt (1, 2)
When calculating TVL it iterates over all the operator delegators and inside it iterates over all the collateral tokens. 
However, the balance of withdrawQueue is incorrectly fetched, specifically this line:
It uses an incorrect index of the outer loop i to access the collateralTokens. i belongs to the operator delegator index, thus the returned value will not represent the real value of the token. For instance, if there is 1 OD and 3 collateral tokens, it will add the balance of the first token 3 times and neglect the other 2 tokens. If there are more ODs than collateral tokens, the the execution will revert (index out of bounds).
This calculation impacts the TVL which is the essential data when calculating mint/redeem and other critical values. A miscalculation in TVL could have devastating results.
A simplified version of the function to showcase that the same token (in this case address(1)) is emitted multiple times and other tokens are untouched:
Change to collateralTokens[j].
Math
jatinj615 (Renzo) confirmed and commented:
Renzo mitigated
Status: Mitigation confirmed. Full details in reports from 0xCiphky, grearlake, Fassi_Security, Bauchibred, and LessDupes.
