Right now there are only two projects which have a 24 hour turnaround: the Faah2 beta and the Fahh2 Async work units. If you don't bother them, they should pre-empt the other units you have running.
Cheers

Plain and simple, scheduling is in FIFO order, until the scheduler determines a tasks needs running EDF to make it to the deadline.

FIFO = First in first out
EDF = Earliest Deadline First

The mentioned statistical app level learning about runtime has been hobbled by WCG with the Don't Use DCF control, so it kind of became task header flops divided by client performance benchmark to get the basic estimated runtime.

DCF = Duration Correction Factor

There's something funny going on when receiving a new task. At times all cached tasks TTC for the specific science will align with the TTC of the newly received task, actually causing scheduler confusion when that change is substantial and running a larger buffer.

TTC =Time To Complete.

So much for injecting some `did you really want to know?' So just let it alone. Interfering will just not help.

As to the core question as posed in the title: Why? Because if you had multiple projects say WCG and Einstein, the project with the shortest deadline would almost always elbow to the front, particularly when buffering more work. FIFO gives everyone equal chance... like standing in queue at Wendy's... you'd be upset if someone barges in later and ihops in front of you. You'd be getting in battle royal if Einstein finds out that WCG runs shorter deadline, so Albert could decide to then too shorter it's deadline.

Of course it's too simple (Way over The Donald's head), projects have weights and usage history, so eventually if Einstein would 'steel' too much resource over time, BOINC would tell them off, go back in the queue and wait your turn which translates to: Stops fetching work from them and tell them to take a hike until the balance has returned. Within WCG that rule does not apply, simple FIFO is the first order and only when there's a deadline situation will EDF kick in and the scheduling is adjusted.

When the WCG server doesn't see your result in time (passing the deadline), your WU will be set to the state 'No Reply' (yet). You will still have some time left to turn in your result then, since a new WU-copy will be sent to a so-called 'wingman'. When your wingman returns his result, which of course takes time, your result should be in, too. So you probably have enough time and get credited, too.

I've only gotten a few HST WUs even though it's the only project I've selected. What's the trick to getting them to flow???

Also, it seems so strange the way WCG switches between its projects. I'll have projects within minutes of finishing and it'll switch. Sometimes it looks like it switched from SCC to FAH but mostly it seems random switching from an almost done OET to a new OET.
It makes it weird when I have a non-WGC project, e.g. POGS, and there's several; WGC projects, e.g. SCC, OET, FAH, & HST. The default Switch Between Projects in BOINC is 60 minutes. It seems that maybe in my example BOINC treats my WUs as coming from 5 projects. But then POGS seems to take forever to finish which seems to have been explained in posts above.
I think it would be most efficient if having started a WGC WU it keeps going until it's done. Should I change Switch Between Projects to 3600 hours???