These failures started happening in the 2007.04.12 nightly.
Here are my analysis report entries:

New nsk.quick_jdi failures (from 2007.04.12)
* nsk/jdi/ObjectReference/waitingThreads/waitingthreads002
        This test failed the following assertion:

            Internal Error (src/share/vm/runtime/thread.cpp:3632)
            Error: assert(count == i,"Did not find the expected number
            of pending Java threads")

        on Linux AMD64 Server VM (machine jck-amd1). This assertion
        also failed with nsk/jvmti/scenarios/contention/TC04/tc04t001
        below.

New nsk.jvmti failures (from 2007.04.07)
* nsk/jvmti/scenarios/contention/TC04/tc04t001
        This test failed due to the wrong # of MonitorContendedEnter
        and MonitorContendedEntered events on Solaris SPARC-64 Server
        VM (machine aditi).

        Update: Serguei believes this failure is due to a test bug:
            6544714 4/4 nsk/jvmti/scenarios/contention/TC04/tc04t001:
                        unsync'ed update of agent's static counters

        Update: In the 2007.04.12 nightly, this test failed the
            following assertion:

                Internal Error (src/share/vm/runtime/thread.cpp:3632)
                Error: assert(count == i,"Did not find the expected number
                of pending Java threads")

            on Linux AMD64 Server VM (machine john), Solaris AMD64
            Server VM (machine vm-x4100-01), Solaris SPARC-64 Server VM
            (machine jtg-s129), Linux IA32 Client VM (machine
            opteron003), Linux IA32 Server VM (machine pacmanjr),
            Solaris SPARC Client VM (machine barkhesh), Solaris SPARC
            Server VM (machine kalieda), Solaris X86 Client VM (machine
            tattoinevm), and Solaris X86 Server VM (machine colfax004)