I think the rationale here is insufficient to justify a removal.
> However, it appears that to the best of my knowledge all hardware
> vendors of SmartLink chips, modems, or whitelabel ODMs are all
> out of business at least as far back as 2005.
Lack of currently-produced hardware is not an argument for ceasing to support existing hardware which may still be in use. (This would not, for example, be a rationale for disabling any of the many drivers in the linux package for hardware that is no longer produced.)
> sl-modem also does not pass the cat-video test - it is unlikely that even
> if one has sl-modem working one can open Ubuntu and watch a cat
> video online.
I don't know what evidence you have to support this conclusion.
> sl-modem-daemon only builds on i386, and has no amd64 userspace support.
Yes, but the i386 package is still installable on amd64, so this doesn't seem to matter for users.
> the kernel driver is not upstream and thus requires constant patching
> by our kernel team to keep building.
"constant patching": there has been exactly one round of patching by the kernel team to fix a build failure, in 2019.
Note that this package was removed once before, but came back via Debian (bug #1650379).
If there were pointers to concrete evidence that this is a maintenance burden for the kernel team, I would accept that as a removal rationale.
I think the rationale here is insufficient to justify a removal.
> However, it appears that to the best of my knowledge all hardware
> vendors of SmartLink chips, modems, or whitelabel ODMs are all
> out of business at least as far back as 2005.
Lack of currently-produced hardware is not an argument for ceasing to support existing hardware which may still be in use. (This would not, for example, be a rationale for disabling any of the many drivers in the linux package for hardware that is no longer produced.)
> sl-modem also does not pass the cat-video test - it is unlikely that even
> if one has sl-modem working one can open Ubuntu and watch a cat
> video online.
I don't know what evidence you have to support this conclusion.
> sl-modem-daemon only builds on i386, and has no amd64 userspace support.
Yes, but the i386 package is still installable on amd64, so this doesn't seem to matter for users.
> the kernel driver is not upstream and thus requires constant patching
> by our kernel team to keep building.
"constant patching": there has been exactly one round of patching by the kernel team to fix a build failure, in 2019.
Note that this package was removed once before, but came back via Debian (bug #1650379).
If there were pointers to concrete evidence that this is a maintenance burden for the kernel team, I would accept that as a removal rationale.