In a previous career as a hardware engineer, I spent two and a half years leading three different projects that ended up being cancelled because marketing realized that they couldn't sell them. These were line cards for an early (2002) racked IP-based DSLAM.

Memory is hazy, and I may get some of this wrong because I left the industry a while back.

One was a multi-port T1/E1 interface card that provided -48V line power to downstream repeaters or CPE. Think PoE but over T1/E1 interfaces at telco voltages and ruggedized to sit in an outdoor metal cabinet with no fans. All components were rated -40C to +85C. I am glad it was cancelled. It was going to be a safety and regulatory certification nightmare.

Another was a multi-port DS3 interface card that did circuit emulation over Ethernet. There were no off-the-shelf ICs that could do everything they wanted. So, we ended up with 4 very expensive FPGAs on the board. This one went into second prototype stage before being cancelled. I'd guess, $200k spent just on prototype hardware.

I can't remember the third project. At that stage I was jokingly known as some sort of project killer.