[adhusson]

withdraw to vault works through deallocate, alternative to #1215

can't withdraw through forceDeallocate but it's not really needed

Addresses Sherlock #18, Sherlock #21 and Sherlock #32.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-31T14:00:15.001010Z	42fcacb	New commits
🔒 Security Review	✅ Completed 2026-10-25T23:10:00.540810Z	7c1b5a7	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[albert-llimos]

This is much better than #1215 imo.

[bhargavbh]

Noticed a slight concern with PA compatibility when thinking about the certora spec. Could be ignored if we assume vaults allocating to Blue and midnight don't support Blue PA.

Before with withdrawToVault:
a user could call withdrawToVault(0) to update the caps and then invoke the blue PA for reallocations. Mainly for collateral caps (shared between blue and midnight).

After this PR:
its no longer possible for someone using Blue PA to update caps and reallocate. The Blue PA would have the authority, since its an allocator, but no entrypoint that calls midnight's deallocate.

[adhusson]

if needed we can add back a public refreshAllocations() that calls withdrawtovault(0) & deallocates