This test is reproducible but may be easiest to reproduce with the help
of Shashank Shekhar. This page:

http://javaweb/~shekhar/serverHotspotVM/1.4/b62/jdk14-b62_servVsClientVM_simp_conn/index.htm

includes performance data and the result of a fastdebug PrintOpto run.

#
# HotSpot Virtual Machine Error, assertion failure
# Please report this error at
# http://java.sun.com/cgi-bin/bugreport.cgi
#
# assert(holder_klass->is_interface() == (bc() == Bytecodes::_invokeinterface), "must match bc")
#
# Error ID: /net/jano/export/disk05/hotspot/auto/ws_jano/fpt/src/share/vm/opto/doCall.cpp, 216 [ Patched ]

The fastdebug run was motivated by performance measurements.
Like earlier tests with Ladybird, there is a mysterious performance loss
running the server vs. the client. For unknown reasons it changes
character with running with the T2 libthread, with the server eventually
getting faster.

This bug documents both the assert and the performance loss.
It is conjectured that the performance loss is a bailout as was happening
with Ladybird; the assert may be an unrelated problem.