To be more concrete... It isn't acceptable for a service to be running from a years-old build, since it wouldn't pick up any security fixes.
Googles infrastructure evolves reasonably quickly - so most reasonably large services will need quite a few code changes every year just to keep up with the fact that internal database type X is going away and everyone needs to migrate to database Y which might have subtly different semantics requiring bits of your service be rewritten.
There is also a gradual flow of legal/privacy mandates (eg. as of Y date, IP addresses must not flow through technology Z because we now have to consider them personal information)
At some point, one of these migrations requires a lot of work, and there aren't enough humans really keen on doing the work, so it gets shut down.
Management could assign workers to do the work, but generally maintaining stuff isn't good for the career path, so few volunteer for big but non-impactful projects like this. As soon as the 'founders' of a project have moved on, it is vulnerable to this kind of shutdown.
Googles infrastructure evolves reasonably quickly - so most reasonably large services will need quite a few code changes every year just to keep up with the fact that internal database type X is going away and everyone needs to migrate to database Y which might have subtly different semantics requiring bits of your service be rewritten.
There is also a gradual flow of legal/privacy mandates (eg. as of Y date, IP addresses must not flow through technology Z because we now have to consider them personal information)
At some point, one of these migrations requires a lot of work, and there aren't enough humans really keen on doing the work, so it gets shut down.
Management could assign workers to do the work, but generally maintaining stuff isn't good for the career path, so few volunteer for big but non-impactful projects like this. As soon as the 'founders' of a project have moved on, it is vulnerable to this kind of shutdown.