Adamw

Yep, please do a short log and attach a copy of the tune.  Here's how:  https://youtu.be/_P1LRANeO4A
 

It is not your lambda probe location that is causing the sensor life issue, it is the curse called innovate.  The lsu4.9 is rated for something like 930degC continuous egt, so temp in a NA exhaust isn’t going to bother it.    Innovate use the sensor in a completely different way electronically than they were ever designed to be, as soon as the sensor response slows down a little their measurement strategy which requires accurate signal timing can no longer work, so you just get the infamous error 8 or error 2.   
Keep your “failed” sensors as 99% of the time they won’t be and will work perfectly fine in any other controller that is not innovate.
 

Our CAN lambda uses the Bosch CJ135 chipset which is their most advanced digital LSU controller, it performs many more error and signal validation tests than most other performance aftermarket controllers as well as some less common advanced functionality for sensor life improvement such as dewpoint detection of the exhaust gas for the correct heating strategy at startup.  It requires a stable clean power supply with adequate current for many of the tests to be effective.  For performance engines which may run CLL at full load, often coupled with a fairly extreme environment for the sensors such as aggressive limiter bangs, fuel additives, silicone poisoning etc, we think it is an advantage to have as many of these validations functional so CLL can be disabled and warn the driver when the signal is suspect.  Yes you can add a capacitor as a fudge in cases where the errors are being falsely reported due to excessive inductive noise or voltage drops, if a clean source cant be found, but this does potentially weaken the ability to detect real errors at the same time.  
There was a small hardware revision to the CAN lambda about the middle of last year in an effort to allow a bit more supply noise.  

You will typically see small/short spikes of error accumulator during heavy foot work - maybe up to about 15, maybe 20 worst case and it should recover to zero within a 10th of a sec or so.  
Log I just pulled from my car below, the worst TP error spikes reach a count of 6.  You can see E-throttle target, TPS sub and TPS main barely deviate from each other at all, you cant even see the green or yellow TPS traces...
  

Change APS (Sub) 100% setting to 100% and change APS (Main) closed voltage to 0.64V should solve it.

Our CAN lambda uses the Bosch CJ135 chipset which is their most advanced digital LSU controller, it performs many more error and signal validation tests than most other performance aftermarket controllers as well as some less common advanced functionality for sensor life improvement such as dewpoint detection of the exhaust gas for the correct heating strategy at startup.  It requires a stable clean power supply with adequate current for many of the tests to be effective.  For performance engines which may run CLL at full load, often coupled with a fairly extreme environment for the sensors such as aggressive limiter bangs, fuel additives, silicone poisoning etc, we think it is an advantage to have as many of these validations functional so CLL can be disabled and warn the driver when the signal is suspect.  Yes you can add a capacitor as a fudge in cases where the errors are being falsely reported due to excessive inductive noise or voltage drops, if a clean source cant be found, but this does potentially weaken the ability to detect real errors at the same time.  
There was a small hardware revision to the CAN lambda about the middle of last year in an effort to allow a bit more supply noise.  

Jon, 
I just replied to your email a short while ago assuming it is the same car we were talking about.  Im 99% sure this is the brushes bouncing off the commutator from vibration. 

Yep there is definitely still a significant variation between sub and main in that log.  I would say most likely it is some variation on the 5V supply or sensor ground as this is common to both sensors.  Since the sensor outputs work in opposite directions, if the supply voltage drops a little, one TPS reads higher and the other will read lower. 
The 5V output in the log shows spot on 5.00 the whole time.  If it were me, probably the next thing I would do is backprobe the 5V and gnd at the throttle body with a multimeter to confirm it is measuring a constant 5.00 there.  Preferably with some long leads so you can drive around a bit with the voltmeter connected.   
The stall is because the throttle position is reading wrong, the throttle blade will sometimes be more closed than what it should be.  For example in your 2nd log at time 4:30 the TP is at 2.8%, over the next few minutes of constant idle the TP drifts all the way to 4.4% to keep the RPM the same, then over the next few minutes it comes back down to 2.9%.  TPS sub does the exact opposite. 

It definitely happens at idle - the pic in my post above shows 6 mins of sitting still idling with the TPS signals diverting away from each other then coming back.  

The location of the cam tooth is the only thing that can change the timing by 360.  The trigger offset is the angular distance from TDC #1 compression stroke to the first crank tooth after that cam tooth passes its sensor.  For the offset to change by 360 it means the cam tooth or cam sensor has rotated by 180 cam degrees in relation to the cam lobes compared to the previous offset.     

It should be fixed now I think.  

It should be fixed now I think.  

You need some way to get extra air in, typically a throttle kicker solenoid or E-throttle.  The idle valve alone wont flow enough.  You will also need a pedal box or vacuum pump for the brake booster as you will have no engine vacuum when antilag is active.   

It is being worked on but was not something we could fix in-house.  There are specialists working on it now but I have no idea how long it may take to correct.  

You can use the CANTEE and CANJST for both power and CAN.  Just be aware AEM use a different pinout in their 4pin DTM connector if your wideband came prefitted with a DTM4 on it.  
The CAN port power supply for the ecu's with the 4 pin JST such as the S2000 is rated at 3A continuous and is protected with self-resetting fuses. It will power most widebands no worries. 
The AEM bitrate is fixed at 500K.  The Link CAN gauge can be set to 500K though so they can run on the same bus.  There is one gotcha at the moment - the bitrate setting is only available in the Android gaugeart app, last I heard it wasn't in the iPhone app yet.   So you will need an Android to do the initial gauge setup.  

You can use the CANTEE and CANJST for both power and CAN.  Just be aware AEM use a different pinout in their 4pin DTM connector if your wideband came prefitted with a DTM4 on it.  
The CAN port power supply for the ecu's with the 4 pin JST such as the S2000 is rated at 3A continuous and is protected with self-resetting fuses. It will power most widebands no worries. 
The AEM bitrate is fixed at 500K.  The Link CAN gauge can be set to 500K though so they can run on the same bus.  There is one gotcha at the moment - the bitrate setting is only available in the Android gaugeart app, last I heard it wasn't in the iPhone app yet.   So you will need an Android to do the initial gauge setup.  

You can use the CANTEE and CANJST for both power and CAN.  Just be aware AEM use a different pinout in their 4pin DTM connector if your wideband came prefitted with a DTM4 on it.  
The CAN port power supply for the ecu's with the 4 pin JST such as the S2000 is rated at 3A continuous and is protected with self-resetting fuses. It will power most widebands no worries. 
The AEM bitrate is fixed at 500K.  The Link CAN gauge can be set to 500K though so they can run on the same bus.  There is one gotcha at the moment - the bitrate setting is only available in the Android gaugeart app, last I heard it wasn't in the iPhone app yet.   So you will need an Android to do the initial gauge setup.  

Yes you should be able to do something to hold the valves open.  
The factory strategy only gives a short burst to open them at startup, then a continuous but low duty cycle signal to hold them open anytime the RPM goes above 4000RPM.  I suspect if you power them all the time the motors will overheat.  
Set up like below should work.  Aux 7 set to high side, on all the time so the motors have a 12V supply whenever ign is on.  Aux 8 set to lowside drive, GP PWM.  When ignition is first turned on we give it 14% DC for 7 seconds.  Whenever RPM goes above 4000 we give it constant 14%DC.  This is similar to how the factory system works when the engine is warm (the y axis on the GP PWM table is Engine running time).

Thats a weird one, it looks like you are doing it correctly but its a little odd how it looks like coms drops out when you click capture.  If you click capture without the engine cranking, do you then get two lines drawn?  
Is this actually a windows PC, not a mac running a VM or something?  And is PC link installed in the default C:\Link G4X\PCLink G4X directory?
 

Something like below is what I would use for an engine like that.

You will need to look up the voltage required for the sensor in the distributor as they all vary.  Often the 8V output from the ecu is convenient.  The signal needs to be connected to trigger 2.  

Something like below is what I would use for an engine like that.

Can you give us a log in closed loop and a copy of the tune.

The idle control mostly looks like it is doing the right things now.  
I suspect your stall issue is possibly fuel related.  Your normal warm idle injector PW is about 0.75-0.9ms in areas of the log where it is idling well.  In all the places the engine stalled the PW dropped down under 0.5ms just on entry to idle - in a couple of places as low as 0.3ms.  I suspect some of your injectors may not even open down at that sort of PW.  In some cases that very short PW was due to reaching very high vacuum in overrun, in others it was the CLL pulling fuel out the whole way down towards idle.  
Try the changes in orange below to see if that improves the stalling problem.
The rev hang is due to excessive offset and base position.  For normal warm idle your TP sits at about 1.7%. In the places with the rev hangs the throttle is sitting at 3.6%.  It looks like the base position for 80°C and above needs to come down to about 1.8%, possibly some of the offsets need to come down a bit too.  
But, lets not change to much at once, try the changes for the stalling issue first before messing with the hang. 

Yeah the view lock shouldnt prevent the table from opening a new view pane as far as I know.   Some parent/child thing is not quite right there by the looks.  It would be pretty rare to lock the cal settings on a page so I suspect it has just never been noticed.
I will report it to the software guys next week.