Is this actually causing any issues? Or just seems like extra spam?

I've been debating declaring these targets 100% of the time, just like the _types target is now always exposed. That way the rule consistently outputs the same labels and the user invoking or depending on a target doesn't need to know if a transpiler is being used etc.

Is this actually causing any issues? Or just seems like extra spam?

I've been debating declaring these targets 100% of the time, just like the _types target is now always exposed. That way the rule consistently outputs the same labels and the user invoking or depending on a target doesn't need to know if a transpiler is being used etc.

I think this is more likely to be extra spam – we only encountered the test failure when building on Windows or under QEMU.
On Windows there's a somewhat meaningful error message (ERROR(tools/test/windows/tw.cc:1302) ERROR: src/main/native/windows/process.cc(202): CreateProcessW("C:\tmp\hs7imhxj\execroot\workerd\bazel-out\x64_windows-opt\bin\src\workerd\api\http-test@ts_project_typecheck_test.sh.runfiles\workerd\src\workerd\api\http-test@ts_project_typecheck_test.sh"): %1 is not a valid Win32 application.) that suggests that the test is not being invoked correctly. That could be an issue with our build environment or an issue with how build_test works but not necessarily a problem with the test itself. I don't have access to a Windows environment so it's hard to tell for sure.

The approach in #721 looks most sensible to me (in terms of not having redundant targets being defined, we always use the default transpiler) but I also get that there's value in defining targets whether a custom transpiler is being used or not.