Confused

The "why" is because it's marginally easier for them to have a single analog pin than figure out how CANBUS works, it's a relatively new technology (introduced about 20 years ago), so many "tuners" haven't heard of it or know how use it
When CANBUS is available for anything, especially lambda, this is substantially better - the controller has the data digitally, why would you want to convert it to a 0-5v output to then convert it back to digital data - it's an unnecessary conversion, and one that's liable to interference and offsets, causing inaccuracies.
Wire it to CANBUS, and be safer in the knowledge that what the controller/gauge is reporting is what the ECU is seeing.

I have only personally tested a box A TC2 and a box A TC4, so im making an educated guess for the ID for the box b.  
TC2 Box A uses 712 & 714.  TC4 Box A uses 716 & 718.
So I think the most likely for box b is 717 & 719.  
You map attached is set up that way.  The probes connected to box A will be received into the ECU as CAN TC Cyl 1-4.
The probes connected to box B will be received into the ECU as CAN TC Cyl 5-8.
 
If the CAN TC Cyl 5-8 runtimes stay at zero then try changing the receive user stream 2 ID to 715.  
Tune with CAN settings 2.pclr

I sold mine to some poor unsuspecting soul and bought a Link CAN Lambda and a 52m GaugeART CAN gauge to fill the hole.
Now I can display heaps of other parameters instead of just the AFR readout.

If you give us a log showing the problem occurring and a copy of the tune then we might be able to help, otherwise the best I can do is consult the magic 8 ball at work tomorrow. 

Yes, the 5v sensor supply & sensor ground can be shared between multiple sensors.

I would set up a new blank page in PC Link with a big digital gauge displaying AN Temp 3 raw voltage, sit the laptop somewhere you can see it then go around the loom and bend/wiggle/yank/stretch anything to do with the oil temp wiring - including where there may be ground splices and especially around the sensor connector and terminals etc.  That will usually find if there is a bad connection anywhere.  If that doesnt find anything then I would probably look at swapping the sensor next.

Judging by what they told you, it looks like they were telling you the ECU can’t ‘power’ a 12V fan, which is true. It just supplies a signal to trigger the relay. 
 

You must power the relay's coils from a switched 12v feed.

You just need to change the Start Position.

With innovate controllers the sensor usually hasnt failed, they just refuse to work when the frequency response gets slower.  If you get error 2 or error 8 the sensor is likely still perfectly fine and will work with any other wideband controller, if you haven't thrown away already, save them as spares for when you have a decent controller.   
Innovate determine oxygen content by driving the sensor completely differently than they were ever designed to be, rather than measuring the actual pump current required to keep the nerst cell near stoich they instead drive the pump at full current all the time, continuously bouncing current backward and forwards, then use the resulting duty cycle to derive how rich or lean it was.  This can give very fast response and works fine when the sensor is new, clean, and its temperature is relatively stable, but age, temperature and deposits effect the timing a lot so the controller then justs reports the sensor has failed if it takes too long to see the response it expects. 
    

One of the biggest killers of the Bosch LSU 4.9 sensors is thermal shock - sitting with the key on and o2 getting hot on a cold motor - then motor starts and splashes cold fuel and water onto the sensor.  The way to avoid this is to not have the wideband power up until after the engine is running. 

The Link CAN Lambda does this via built in function over CAN.  I think the Spartan 3 also does this over CAN, but I've never used the Spartan 3.
I use the AEM X-Series in-line controller which has CAN or Analog output if I'm not using the Link CAN Lambda.  On the plus side, it's about half the cost of the Link unit, but on the negative side I usually have to add a relay to give the ecu control of when to turn the controller on/off via a spare output. 

I believe the Innovate Lambda controller instructions should read as follows:

Step 1: Remove from vehicle
Step 2: Throw in trash
Step 3: Install any other lambda controller
Kidding/not kidding.  If you don't have a viable way to replace the lambda controller then it is what it is, but innovate controllers are known to have issues with failing sensors extremely quickly.  I would keep spare sensors on hand.

Your screenshot(s) are telling you what you need to know.
TPS Sub - An Volt 7 - is currently showing 4.21v, and you're telling it to throw an error if it's above 4.00v - which it is - you're also telling it to not throw a fault if the input goes to 0.00v, which is not good either.
TO GET IT CALIBRATED ONLY - set the Error Low to 0.05v and Error High to 4.95v for all 4 sources (TPS Main, TPS Sub, APS Main, APS Sub), and then look at what the minimum and maximum voltages are (the Open and Closed voltages that are auto-populated once it's calibrated).
 
For example, if TPS Main spans from 0.55v to 4.32v, then I would set my Error Low to something like 0.4v, and Error High to something like 4.4v - this is just outside of the range of expected movements, but if a fault does occur, it'll pick it up.
 
I strongly suggest, if you haven't already, to read the Help within the application a bit more thoroughly - it's exceptionally well written, and unlike some other manuals/help I've seen from competitors, it's not just a reference guide that assumes you know how stuff already works, it actually explains what and why very well. It's the very best Help/Manual I've read in a very long time!

Your screenshot(s) are telling you what you need to know.
TPS Sub - An Volt 7 - is currently showing 4.21v, and you're telling it to throw an error if it's above 4.00v - which it is - you're also telling it to not throw a fault if the input goes to 0.00v, which is not good either.
TO GET IT CALIBRATED ONLY - set the Error Low to 0.05v and Error High to 4.95v for all 4 sources (TPS Main, TPS Sub, APS Main, APS Sub), and then look at what the minimum and maximum voltages are (the Open and Closed voltages that are auto-populated once it's calibrated).
 
For example, if TPS Main spans from 0.55v to 4.32v, then I would set my Error Low to something like 0.4v, and Error High to something like 4.4v - this is just outside of the range of expected movements, but if a fault does occur, it'll pick it up.
 
I strongly suggest, if you haven't already, to read the Help within the application a bit more thoroughly - it's exceptionally well written, and unlike some other manuals/help I've seen from competitors, it's not just a reference guide that assumes you know how stuff already works, it actually explains what and why very well. It's the very best Help/Manual I've read in a very long time!

The CAN gages on the market are pretty cool. Link sells ones that appears to be a private-labeled version of the GageArt unit. BTI sells another.
Bottom line, it can be configured to display anything the ECU is aware of. It can show one, two or four channels at a time. You can configure multiple "pages" and cycle through those pages with the touch of a button. Very nice.

It's the White (0-5v Output) wire you want to connect to an Analog Volt input on the ECU.
Then create a Linear Cal for Lambda and make it 0v = 0.683 and 4.99v = 1.365.
I have to warn you though... This Lambda controller is garbage.
I used one for awhile and it gave terrible readings that changed day to day because it doesn't have a dedicated Sensor Ground wire.
It's good for a "close enough" overview of if you're "lean", "rich" or "about on target"... but absolutely useless to use for accurate Closed Loop control.

Pretty easy to figure out a TPS without the pinout.
Using a multi-meter on ohms - with the power off; probe 2 pins until you find the 2 where the reading doesn't change when you open the throttle.
They will be Sensor 5v and Sensor GND. The final pin is the Signal.
Leave the black lead on one those pins and move the red lead to the final pin.
If the reading starts low and increases as you open the throttle - the black lead is on Sensor GND.
If the reading starts high and decreases as you open the throttle - the black lead is on Sensor 5v.

Unfortunately E85 isn't available here, but every post I've seen here with multi fuel is using multiple fuel tables...
 
One thing that I've just noticed, you mention you tuned the VE table on E10, but in the blend table you're starting from 0% ethanol with 0% blend, and at E10 (where you tuned) you're already saying to blend 8% towards the second fuel - if the value in the E0 and E10 cells is 0%, then ramping up to 100% blend at E85 might give you a closer result?
 
As I said, unfortunately no first hand experience, but I've been reading posts here daily for 3 years now!

You don't have to connect up an output from the sensor to the ECU if you don't wish to.
I have a combined sensor in my fuel pressure regulator, but I'm only using the Pressure side, so I have connected 5v, sensor ground, and the pressure output to the ECU.
So, yes, that's correct

Yes they are completely sealed, the internals are fully potted in epoxy and the anodised aluminium seems to stand up well to hasrsh environments.  

A number of pin functions can be changed through software, but not Injectors and Ignition drivers that are being used for Injection and Ignition. So, Injectors must be pinned to Injector 1-X outputs, and Ignition must be pinned to Ignition 1-X outputs, and you can't skip over any (i.e. you can't use Ignition 1-3, then 5-7, you must use Ignition 1-6)
 
Whenever you use a vehicle specific ECU in a different vehicle, it's your responsibility to match the car's wiring to the ECU. 

It's power ground, so a regular ground to the chassis.

I vote for CAN connectivity.  

Dead time has the largest affect at idle or in other areas of the map where the effective pulse width or duty cycle is very low as the dead time in these cases will often be much higher than the calculated effective pulse width.  Actual pulse width commanded to the injector is the addition of injector response time aka dead time and calculated effective pulse width.
Usually provided data is not 100% correct unless it was set up with the same ecu (so it has the same driver hardware).  It is very critical also to make sure you have the correct engine displacement data, and injector flow value entered.  You also need to be sure you're setting the dead time for the differential fuel pressure the engine is operating at. 
If you're unsure if your deadtime is correct, and if you want to check it - do the following after warming up the engine fully and having no more warm up enrichment active (or any other active fuel trims ideally):

1. tune idle to run right on lambda target with closed loop off or as close as you can. (for example 1.000 lambda)
2. Change lambda target to 0.9 lambda and see what value the afr goes to.  If dead time is correct - the new lambda value will match target.
At this point if it is not hitting the new target the dead time needs to be corrected.  The following is what I do, and may not be the best way, but has worked well for me:
A. If when changing from 1.0 lambda to 0.9 lambda target the AFR undershoots - does not go all the way down to 0.9 - you have too much dead time and not enough resolution on the VE table.  Reduce dead time and retune the VE table to match target.

B. If when changing from 1.0 lambda to 0.9 lambda target the AFR overshoots - goes past or below 0.9 - you have too little dead time and too much contribution from the VE table.  Increase dead time and retune the VE table to match target.  
Rinse and repeat.
Ideally you will do this at multiple voltages to generate a correct curve, but being able to apply multiple voltages to a car's injectors, ecu, and ignition while it's running isn't something everyone has the ability to do, and in the real world you're usually going to be operating at roughly the same voltage plus or minus a volt 99% of the time, so it's certainly critical at a minimum to get the dead time right at nominal operating voltage.  If you have an adjustable regulator you can also perform this same test at multiple differential pressures to get a good 3D dead time table, it just takes time.
 
Having correct dead time makes all the tuning go quicker afterwards as all changes that are predicted to be needed based on AFR readings will be more accuately applied, as well as all fuel calculations and corrections the ecu is using will be applied to the actual effective injection pulse.
 

I would say you have just kissed the RPM limit.  You just dont see all the detail due to the logging rate.  In the next firmware the limit capture for logging has been improved so that the status and limit flag word will stay active for a minimum time so it is always captured even in a slow log.