I've had recent dealings with two sides of the 'embedded Linux' vendor participation scene:
1) One vendor was primarily interested in using the Linux kernel as the basis for a proprietary development and application layer. It just so happened that they were the sole source of support for this particular flavor of embedded processor at the time, so (our mistake and theirs) we got in line to get the modified kernel sources, rather than take the porting effort on ourselves (after all, we wanted to concentrate on our application, not do a kernel port). Since they were so focused on their product, rather than the kernel aspect of what they were doing, it had forked rather badly from the main kernel. When the company suddenly closed its doors, we managed to get the GPLed stuff from them (barely). It was more or less useable in its then current form, but completely unsuitable for submission to the kernel tree of interest. In the meantime, the target platform evolved, making the port even more obsolete.
2) At this point, the chip supplier stepped in, and decided that they would partner with someone else to finish the port. We waited on the sidelines for some time, until the new partner was ready to throw the GPLed code over the fence. A further annoyance was the existence of substantial sample source code freely available from the chip supplier, but with a non-GPL friendly license attached (redistribution of derived works allowed only in binary form). Talking about re-licensing or dual-licensing the code (they also had closed source customers that wouldn't be able to use GPL-only sample sources) seemed to just die out once it reached upper management levels. I also tried to convince the chip supplier to set up a communications center (weblog, sourceforge project, email list, whatever) so that all the orphaned customers of the first failed port could help each other and exchange info. This also fell on deaf ears.
In essence, we lost about six months with all the machinations, and the project was eventually shelved, just when we were about to finally get started in earnest.
With 20/20 hindsight, if we had spent half of that time getting our own platform up and running, we would have been much farther ahead of the game, and not dependent on outside commercial interests.