Given enough time or numbers, even the lowliest mooks will roll 20s.

...or you'll encounter someone who ain't just a lowly mook, even if they're also not the big bad. :)

In 3.5, that doesn't necessarily matter, because skill checks don't auto succeed or auto fail on 20's and 1's, without house rules saying so. It is entirely possible to have a bonus so much higher than someone that even if you roll your absolute worst and they roll their best, they can't spot you.

This touches on a thing me and some friends try and crack down on. 'Crowdsourcing success.'
In this case, say, it makes sense. Everyone is looking for Zoro, so the thousand monkies rule is in effect. A lot of people searching for Zoro have a good chance that one of them is actually going to beat the roll.

But this becomes a problem when PCs try to pull it on tasks that it doesn't make sense for. The big example being the party came across a computer in a room they were searching. One of the players asked to investigate the PC, followed by the rest of the party chorusing the same. This would mean, say, 5 people rolling against the one problem to try and have the highest chance of succeeding... even though the actual task, say 30 minutes of searching through files, could NOT be done by all of them at once.

I mean, they could.

More eyes and discussion between them as they look through files.

That's when you just give some kind of marginal bonus to the single check, imo. Having people backseat direct you when you're on a computer, or vice versa, is awfully inefficient and frustrating.

Yes, having two people who know what they're doing *might* make things faster. Having three or more people, even if they are all qualified, will inevitably make such a task slower.

I'm all for bonuses. Heck, many systems have an 'assist' feature for helping with a roll. But think if you were all in the one car and asked for a drive check, everyone rolled.