Some brainstorming:

• What if scripts could provide an estimate of number of connections they will make? or seconds they will sleep? Then NSE could track actual connections and sleep events against that estimate.
• A common case is one script holding things up at the end--for example if the user did --script ftp-* and didn't realize "ftp-brute" was included. When we print the stats at the end, maybe we could check and print the names of the remaining scripts if there are less than 5 or so? Nothing the user can do at that point, but it could be helpful for reporting slow scripts or avoiding them in the future.

Adding to your suggestion, we could have a counter for the number of threads spawned and based on the estimated number of connections that a script makes, we can provide a (somewhat) fair estimate of the time remaining. But I don't know whether it would be possible to estimate the number of connections for all scripts though.

Certainly having an idea of which scripts are the ones holding up the end of the scan would be nice. I've found that using --script-timeout=15m is a good solution. It gives the scripts time to finish on their own, but has hard cutoff if for some reason they get stuck.