I was profiling renaissance's finagle-http for misaligned memory reads and found few cases in String.indexOf intrinsic.

For example ( trp_lam is a perf event for misaligned read faults):

#
# Samples: 20K of event 'trp_lam'
# Event count (approx.): 63238
#
# Overhead Command Shared Object Symbol
# ........ ............... ................. ..................................................
#
    84.40% finagle/netty4- [JIT] tid 21164 [.] int java.lang.String.indexOf(java.lang.String)
    14.75% java [JIT] tid 21164 [.] int java.lang.String.indexOf(java.lang.String)
     0.77% finagle-http-wo [JIT] tid 21164 [.] int java.lang.String.indexOf(java.lang.String)

Later I found this also can be partialy reproduced with StringIndexOf.advancedWithMediumSub jmh test

In some cases, String.IndexOf intrinsic performance could be up to 50x times worse than expected. ( on h/w without misaligned access support)