host's current tasks don't show up on website after merging

That ID was created on 26th April, 17:41:31 UTC.  That's close to 7 days before your initial message about the situation.  It seems to be the original ID, not the new one.

I called it a "companion".  What I meant by that was that there was a second listing in your hosts list with what looked like an identical set of hardware.  Perhaps I should have called it a 'twin brother'.  I don't know if you have a second machine with the same hardware but if you don't then you have managed to get a second host ID for the one physical machine.  It's not really supposed to happen but it does.  It doesn't happen to me so I don't have any clues as to exactly how it happens.  The ID 12775435 was created 2 days later on 28th April, 13:34:39 UTC.  Perhaps you might be able to remember what you were doing at that time that caused the second ID to be created.

By 'deleting' are you referring to the button in BOINC Manager currently called 'Remove' - it used to be called 'Detach' (I think).  I just want to be sure you didn't delete the BOINC tree or something like that.  I've never had to detach and then re-attach but I believe that you are supposed to get back the same host ID if you do.  That seems to be exactly what happened with your "deleting the project", so that's not surprising.  Your duplicate ID was 12775435 and you got that same one again - nothing changed.

Your former ID 12775239 shows 8 tasks - 1 valid, 1 invalid and 6 error.  depending on how long others take to complete the 8 quorums, it could be weeks before they all disappear.  If you merge at that point, everything should successfully end up under the current ID of 12775435.  When I looked, that one showed 177 tasks - none completed, 81 errors and 96 in progress.  Is that machine actually crunching?  It's rather odd that nothing (apart from errors) has been returned.  The 'in progress' are all FGRPB1G GPU tasks.  Is there a problem with the GPU?

I currently have a total of of 587 host IDs under my account.  At the moment I have 108 actively returning work of which 104 have a decent GPU - about 5 or 6 have 2 GPUs.  Around 10 years ago, I decided not to unnecessarily create more new host IDs if I could avoid it.  At that time I had over 200 active machines (CPU only).  I'm not really into nostalgia but I certainly wouldn't lose out on it if I were.  When I decide to reuse an existing host ID, it's very easy to write a short entry in a notebook of the website details as of that time.

In my previous post I gave a link to the most recent machine I've added, the i7-4771 I purchased recently.  It reused the host ID of a Coppermine PIII 933MHz that ran between 2006 and 2010.  By chance, I just stumbled on this old message I wrote in 2008 within which I have a link to that very PIII 933MHz. If you click that link in that old message, you will see today's i7 :-).  What a coincidence! :-).

I chose PCLinuxOS around 2007.  Up to that time I had been attending computer auctions and buying ex-business machines sometimes workstations, sometimes servers.  I came across a large number of machines all the same brand, all with Tualatin Celeron 1300 processors.  Because of the 'Celeron' in the name, most other bidders avoided them like the plague so I had virtually no competition.  They were presented in job lots of 20 at a time with one lot per monthly sale.  Over a 9 month period I ended up with about 8 lots.

I had already had a good experience with Coppermine PIII's and excellent results from Tualatin PIII's in servers.  With a bit of research I discovered that Tualatin Celerons were just as good and were easily overclockable.  People avoided them because of the very poor performance of the early P4 Celerons.  The boxes I was buying all came with Windows XP product codes on them so I started running them under XP.  As the number of hosts grew, management became increasingly difficult so I decided to try Linux.  PCLOS was my first choice and I found it to be a perfect fit for the way I like to do things.  Back in the late 1970s I had used the Bourne shell under BSD Unix and had done quite a bit of shell scripting so the Bash shell under Linux was a joy to use for scripting purposes.  I really liked the KDE 3 desktop so converting everything to PCLOS was a no-brainer for me.  I've since had very brief encounters with other Linux distros but nothing compares with how well PCLOS works for me.  Find what works best for you and stick with it is my advice!! :-).

On the subject of merging 2 copies of the same machine, my understanding is that one doesn't need to wait until the results are purged from the database. The older host number should be merged into the new host number and all tasks in the database should follow along.
When one wants to delete a host record from the database, then one needs to wait until every task associated to the host is purged before that action is allowed.

OK, that explains it.  The GPU was probably running Seti tasks rather than Einstein when I looked.  I didn't realise you were running Seti as well.

All computers are in a single 12m deep x 6.5m wide room on the ground floor of an industrial warehouse I own.  They've been there for the last 7 years.  My space is a small separate area at the back of (and below) a much larger tenancy that is used by a commercial Real Estate business.

There is no air conditioning used for the computers.  I have installed two large industrial fans with suitable ducting, one to bring in fresh outside air and the other to expel heated air.  The airflow is through the computers which are all 'open case' and mounted in pallet racking to allow good ventilation.  It works a treat and despite the sub-tropical climate, I run all year round with very little 'excess heat' problems.  I do have (through a purpose built control script) the ability to pause crunching if the room temperature becomes 'excessive' :-).

As an example of longevity, here is a Q6600 quad core machine I built as a CPU cruncher in 2008.  These days it has an RX 580 but it has been in service 24/7 for more than 10 years and it's still going strong.  I built 6 of them at the time and they're all still running as GPU hosts.

Two years ago, I installed 100 x 340W solar panels on the roof of the building.  I commented on that in this particular message in Mike Hewson's "Sunny Thoughts" thread that he started in the Cafe.  In my next message after that (July 9), I showed a "Site Layout" for the solar panels and several pretty pictures giving graphs of power and energy stats as of mid-winter.  6 months later, at the height of summer, I showed another graph showing the massive increase in production for summer.

The Arch wiki is great.  I use it quite a lot whenever I need good detailed information about stuff I'm not familiar with.

Older GPUs that still have a reasonable crunching capability may well have had a hard life in a gaming rig.  At the very least, expect fan lubrication problems and perhaps dubious BIOS mods or voltage mods to 'improve' gaming performance.  That would be too much of a risk for me to feel comfortable with.  They will also be power hungry and probably quite inefficient for crunching.  You need to choose carefully.

Be careful.  Although PCIe bandwidth with the current Einstein apps, is not as limiting as it used to be, it's still likely to cause problems.  The bandwidth available when using x16 and x8 is certainly quite OK and there doesn't seem to be a huge penalty if using x4.  You should at least hook up a single GPU through an x1 slot and riser (whilst still using your x16 slot(s)) and test for yourself.  It will crunch more slowly and if it seems acceptable, hook up a second or third if you can because each additional connection will subtract from the overall available bandwidth and cause slowdowns.

The bandwidth requirements are quite different to what you have with mining apps and my expectation is that the performance through x1 slots will be rather poor.  You really need to test this for yourself.

Asrock have this mining board that they claim will support 13 GPUs.  I only know about it because I noticed that one of my local computer shops has it on special for just $AUD55.  They quote the RRP of $AUD160.  That gives me a very clear message that either the board is a dud or that mining is in such a decline that the suppliers are desperate to get rid of excess stock while they can.  It's most likely the latter.

The board should have a very decent voltage control system so it should do a very good job of running a single high end GPU in the sole x16 slot.  I might buy one and stick a Radeon VII in it :-).  I do have a spare RX 570 so I could try that first - just in case the Asrock board was a dud after all :-).