Harald,
I believe this is more correctly phrased as "If the mobile node no longer has link connectivity with the old subnet, any packets...."
The obvious, but not always practical, solution to the problem (keeping both links during handover) is impliclity alluded to later, under "FMIPv6" (link preconfiguration), but is not mentioned explicitly anywhere.
"Not always practical" is the operative phrase here. I often hear this solution proposed, unfortunately, to my knowledge, there is only one wireless link protocol that supports multiple handover legs for IP today (Flarion's OFDM protocol) and it is not a particularly big contender commercially. In particular, the "soft handover" approaches of wCDMA and cdma2000 don't work for this purpose, the handover legs are effectively invisible at the IP layer.
And of course the obvious common question: Where are the security requirements going to be developed....? (HMIPv6 introduces an obvious point for man-in-the-middle attacks, which may not even have to be on-link, for instance. So it's not security neutral....)
HMIPv6 has some security currently, not sure if it addresses your particular concern since it's been a while since I looked at the document. The security considerations in the current FMIPv6 draft are rather weak, there are some approaches that have been discussed in the WG, most utilize IPsec in rather cumbersome ways or depend on as-yet incomplete solutions like AAA key exchange. Erik Nordmark mentioned the possibility of deferring specific security solutions for FMIPv6 until and unless the work went from Experimental to Proposed Standard, don't know whether that's a still viable proposition, but I think it may be difficult to come up with anything clean in the short time period during which this WG is expected to exist.
One of the intents of the WG is to get the work published as Experimental, since it has been underway for 3 years now, then sort out integrating the various bits of handover related technology that are being done in various parts of IETF within the IP Mobility Research group in IRTF, filling in holes where necessary and coming up with a clean, integrated handover solution having strong security. That would then be a possible candidate for PS.