No, I expect my scripts to run for a long time with many users. This doesn't preclude optimization; some of it is automatic (new hardware, interpreter library improvements), some of it is well established (using SWIG and C to replace only a tiny piece of the program that must be faster).

In some respects, having long-lived software with lots of users makes speed the least of my concerns, because they're always asking for new features, and those are relatively easy to add to scripts.

And the relationship between software speed and productivity isn't linear, because people multitask. If a program takes 10 seconds to run, I might sit and wait for it to complete, without doing anything else. Whereas, if the program takes a minute, I may decide to switch to another quick task, and then return to see results. In this case, both tasks needed to be done, one took longer but it ate up the "slow" runtime of the program, and was only parallelized because of that long runtime.