Try using Virtualbox from Oracle. If you are using in commercial company, you may have to pay license.
It is a stand alone prgram that wraps the virtual machine.
Virtualbox has a nice intuative user interface.
The virtualisation allows complete control of network interfaces.
It also allows the virtual machine access to the host machines USB port, so you can still use USB devices. You just have to enable the usb in the virtual machine (vm) controls.
Once installed and started, create a new vm, create a disk drive for it (creates a file for the drive image which you can place anywhere on the host filing system) and assign something to boot from eg CD/DVD drive - you need this to install the vm operating system.
Start the vm.
The vm will boot from the boot drive and if it is a WinXP install disk, it will install onto your vm.
Let the WinXP install finish and there you go.
You should be able to boot and shutdown WinXP any time in the vm.
You can even suspend (pause) the vm (do not do it during some disk or usb operation )
Try using Virtualbox from Oracle. If you are using in commercial company, you may have to pay license.
Not quite correct. You can use virtualbox for free unless you use it to offer virtual hosting services.
If you have Windows 10 Pro or Enterprise, you can also use Microsoft's Hyper-V.
How can it be a problem to upgrade legacy well functioning tools to Windows 10?
It's a HUGE problem, exactly analogous to your "how can it be a problem to support legacy HCS12X and HCS08 projects?"Those old tools aren't upgraded, or even maintained. They're just "archived."
I'd lay odds that their source code for CW4.7 doesn't even compile with modern windows10 toolchains. :-( (and if they compile, I'll bet they're rife with bugs of the sort where they assume sizeof(int)=sizeof(void *)=32bits, making it at best problematic to compile for a 64bit environment.)
Are your applications written in C, or assembly? Isn't the HC08 pretty similar to the HCS08 and the 12X similar to the 12Z? Can you get by with magic switches to the CW11 tools, or simply "not using" new instructions? (similar to the way that gcc for 68000 is now sort-of "coldfire minus xxx, minus yyy" ?
How can it be a problem to upgrade legacy well functioning tools to Windows 10?
It's a HUGE problem, exactly analogous to your "how can it be a problem to support legacy HCS12X and HCS08 projects?"Those old tools aren't upgraded, or even maintained. They're just "archived."
I have a handful of applications, larger customer systems,
and smaller tools, all written 20-30 years ago.
They run under Windows 10.
It was not always easy to maintain functionality
with newer OS versions, but I managed, with greater or less effort.
Where there is a will there is often a way.
I have CodeWarrior 6.4 targeting ColdFire on a Windows 7 machine. I use the P&E programmer / debugger through USB.
I tried but was unable to get CW 7.0 to work a long time ago.
I will likely get an "upgrade" to Windows 10 in the next few months, so this topic is timely.
The XP computer previous to this was virtualized just in case, although it has been some time since I used it. I don't recall if all the debugging features worked, but neither do I recall them not working.
IIRC the P&E programmer doesn't always work from a virtual machine using Virtualbox but it has been several years since I have used it. I vaguely recall that it also depended on the machine so the type of USB port may cause problems.
No idea.
What I would suggest is to take every opportunity to tell Freescale/NXP how bad Eclipse is !
Hopefully, if they understand how bad Eclipse IDE is for embedded tools, they will stop using it as the base for their tools and go back to actually developing a decent tool 
You may dislike it, but Eclipse is your best bet in support across many OS versions over the years - we have Eclipse-based software that has survived generations of Windows versions, as well as cross platform (OSX and *nix). Prior to that, the tools used to become quite fragile after the fourth or fifth service pack released by Redmond (2k and XP days) - support for other OSes was not economically viable due to the sheer number of users back then.
Developing an IDE from the ground up requires an army of developers and testers - for most of the companies the return on investment is quite narrow, especially in these days of "everything must be free". Let alone the reduced number of users in other OSes, which is also quite difficult to accomplish with a full custom design, where companies like IAR, Keil, Segger, Altium and others still didn't make the jump.
where companies like IAR, Keil, Segger, Altium and others still didn't make the jump.
You have listed all the companies that ARE successful because the free Eclipse based tools are so dreadful that companies are prepaired to pay lots just to have something that works and is reliable.
where companies like IAR, Keil, Segger, Altium and others still didn't make the jump.
You have listed all the companies that ARE successful because the free Eclipse based tools are so dreadful that companies are prepaired to pay lots just to have something that works and is reliable.
Yeah, sure.
Tell that to the number of developers that migrated to their competition due to the lack of multi-platform support.
Agreed. The vendor provided IDEs generally suck because they are not the vendor's core business. Compared to all the IDEs I have worked with Eclipse is by far the best because it actually helps with navigating through big software projects. And Eclipse also supports more programming languages (think about Python, Lua, VHDL, Verilog) compared to vendor provided microcontroller IDEs so you can have one environment to do everything.
Try using Virtualbox from Oracle
VirtualBox is known to have a serious problem with the USB virtualization. If you use a USB-cable (e.g. a Xilinx cable) that needs to changes its USB endpoints when it uploads its internal firmware, then the virtualizer may or may not catch the event, and if doesn't catch the "new USB endpoint", you will loose the connection, you have to unplug, and try it again.
I suggest you a better virtualizer. Xen, or Fusion, or WmWare.
Try using Virtualbox from Oracle
VirtualBox is known to have a serious problem with the USB virtualization. If you use a USB-cable (e.g. a Xilinx cable) that needs to changes its USB endpoints when it upload its internal firmware, then the virtualizer may or may not catch the event, and if doesn't catch the "new USB endpoint", you will loose the connection, you have to unplug, and try it again.
I suggest you a better virtualizer. Xen, or Fusion, or WmWare.
In my experience, VMware is the best tool for USB virtualization. Not 100% perfect, but still the best.
Perhaps OT: I've been using daily Virtualbox in Win10 host running Ubuntu guest. While running iceprog (from icestorm suite) flashing 104kB (ice40UP5k via FT232H) took 165seconds. I got an advice (icestorm github) to mess with the usb settings in Vbox. USB1.1/USB2.0 took 165secs, luckily the USB3.0 gave me useful 15 seconds. The Biggest issue with the Vbox I still see (for many years actually) is the copy/paste from/to host, still crappy.. Otherwise it works pretty good (I used to use vmware in past).
FT232H
FTDI chips don't need to upload their firmware, so they don't need to show different USB endpoints.
But chips based on 8051 do need to do so! So, Xilinx and Altera cables show to different USB endpoints.
First, the endpoint required them to receive their firmware, which is send from the host to the chip.
Then they reboot, and they show the final endpint, which is the application endpoint, the one used by the Xilinx program "Impact" to program CPLD and FPGA chips.
They say that it will not work for this or that reason.
I find it hard to believe their tools do not work in W10.
Why do you find it hard to believe? Xilinx ISE doesn't run on Win 10 unless you diddle around with some filenames or run it in a sandbox. Fortunately, changing a few filenames more or less solves the problem but it is not supported by Xilinx. They suggest the sandbox.
Things changed between Win 7 and Win 10 and some applications don't run. It's a fact, there's no point in questioning whether it is true.
I searched:
https://html.duckduckgo.com/html?q=HCS08not even realizing it was for the old 68HC08. That's a processor from the '80-ies or so I think?
The first that popped up was an IDE from cosmic. I have no idea about the quality of their tools.
But for that old stuff it seems like a complete revision of the project is in order, especially for > 10k production volumes. Lots of things have progressed in uC land and such an old uC is not quick and may be replaced by a 40ct part. Apparently you use C compiler, and C-code is relatively easy to port to any microcontroller, especially if it's well written and not too heavy on I/O or dependent on special features from a particular microcontroller.
With those old parts the availability of the hardware itself may also become a problem.
I did not read the whole thread, but to me it looks like it's getting time to have a broad view look at the whole project and set out a strategy for it that will work for the next 10+ years.
I myself have a preference to keep things simple.
Just give me a decent C compiler and some header files and startup code.
I really do not like makefiles, but once you've got a project working with GCC and make, you usually have a setup that is relatively stable and reasonably future proof.