Not sure if you've read that (or if I wrote it in a different thread), but Peter and I have continued investigation by checking old data sheets to see what's the guaranteed high-level of the old LS157 parts. We found that the problematic one was way off, so technically, it must be seen as a defect on the component side. The fix that I've found (pull-up resistors) has helped, but it's not clear how long, as the defect may continue to creep further.
Ageing of integrated circuits and even discrete semiconductors is an interesting topic. Not for thie thread, though...