It would be very useful to have pacman pid stored inside /tmp/pacman.lck.

This way, when another pacman instance runs, it can check if the process is still running, even if the process is not called pacman (let's say, someone made an alternate package manager).

in addition it would be good to store which tasks pacman is performing, so that if it is just performing a search you know that you can install a package at the same time, or that if it is doing a sync a search won't be up to date in a while, or may be corrupted. This would also imply making the pacman.lck multiprocess. Together with this a frontend should be created (let's call it pacmanlib), so that you can do "pacmanlib --add-process $PID $STATE" and "pacmanlib --remove-process $PID". pacmanlib would be the multiinstance manager, so, if you want to do "pacmanlib --add-process $PID SYNC", it returns an error message (it doesn't need to be an output on STDERR) if this operation can't be done in this moment because another processes is performing a conflicting task. pacmanlib could also be responsible of checking if the processes that are listed in /tmp/pacman.lck are still running.

pacmanlib would make completely clean the access to the database.

ps: I call it pacmanlib because i think it is a good name for pacman library's frontend.