crypto/cipher:gofips140: TestCTR_AES_multiblock_XORKeyStreamAt failures

[gopherbot]

#!watchflakes
default <- pkg == "crypto/cipher:gofips140" && test ~ `TestCTR_AES_multiblock_XORKeyStreamAt`

Issue created automatically to collect these failures.

Example (log):

=== RUN   TestCTR_AES_multiblock_XORKeyStreamAt/keySize=24
    ctr_aes_test.go:301: multiblock CTR's output does not match generic CTR's output
--- FAIL: TestCTR_AES_multiblock_XORKeyStreamAt/keySize=24 (1.84s)

— watchflakes

[gopherbot]

Found new dashboard test flakes for:

#!watchflakes
default <- pkg == "crypto/cipher:gofips140" && test == "TestCTR_AES_multiblock_XORKeyStreamAt/keySize=24"

2025-04-01 00:06 gotip-darwin-amd64-longtest go@4c32b1cc crypto/cipher:gofips140.TestCTR_AES_multiblock_XORKeyStreamAt/keySize=24 (log)

=== RUN   TestCTR_AES_multiblock_XORKeyStreamAt/keySize=24
    ctr_aes_test.go:301: multiblock CTR's output does not match generic CTR's output
--- FAIL: TestCTR_AES_multiblock_XORKeyStreamAt/keySize=24 (1.84s)

— watchflakes

[FiloSottile]

This looks like memory corruption in gotip-darwin-amd64-longtest like #72020. They are both tests that compare large portions of memory.

Should we replace this builder?

/cc @golang/release