Start Power timing

The B generator comes on line immediately.
You can see its effect because it back feeds out the start Power connector and your battery voltage will immediatel jump above 13v before your hand is even off the Start Key.

There is No hurry to remove start power.
The ECU will internally (Relay) disconnect the START Power shortly after the engine starts.
According to the Operators manual, this happens once the engine reaches 2500 rpm.

You could leave the Start Power engaged for the entire flight without issue.
The real reason for disabling the Start Power is in the event of a crash, when the engine stops as the prop hits the ground, the internal generators also stop, de-energizing the Fuel Pumps.
If the Start Power were still available, The ECU (Relay) would revert to the Start Power source and the Fuel pumps would again become powered.
It would not be be a good idea to have operating fuel pumps if the crash also compromised the integrity of the fuel lines and a fire ensued.

15-60 second would be a reasonable delay time.
As long as the system disables the Start power before you take the runway, that would be OK also.

If the Engine stalls, be sure your system can NOT re-energize Start Power without Pilot Action.
- - -

An "Off-A-B-Both-Start" Key switch is a very Bad idea on a 912is Engine.

I have an "OFF-ON-Start" Switch that I reset to OFF after starting.
I "Could" pull the key and put in in my pocket for the flight, but It is kept in the Switch "Just In Case".

The Lane A, B, and Pump Switches should be separate and independent of the Start Switch.

Fuel Injected engines are wired significantly different form their Carb'd Cousins.

On my AirGyro 915, I often have rough starts where the engine sputters at low RPM for about 5-10 secs before reaching normal idle speed.  Twice, I had Lane lights start flashing with no RPM indications.  I wondered if it might be due to low battery voltage so measured voltage after start across the battery terminals with the start switch on and off.  On=13.7V; Off=12V.  

With those readings, it seems tha battery is not charging with the start switch off.  Does that sound correct?

If either lane light is flashing, The ECU is in engine protect mode and battery charging MAY not happen.
The CAN Buss may also disconnect so no engine data will be displayed either.

The battery voltage should never be below 13.0v when the engine is running.
Is is not unusual for the battery voltage to drop into the 10v range during cranking.
Often the computers will Reset or worse...  Partially Reset.
It is not a bad idea to Reset the lanes manually after a start.  It couldn't hurt.  A OFF-ON, Pause, B OFF-ON.

As instructed, be sure to open the throttle half-way before cranking, especially if the engine is not already warmed up.
Yes, it will roar to life and immediately run in the 3000-4000 rpm range.
Keep your hand on the throttle and pull it back as soon as it starts.

I recently had a start where it appeared to be running on 2 cylinders.  Lots of shaking!!!
I suspect that one of the lanes had dropped out but Not Reset and the other lane thought it was still working and what was really happening was it was firing all 4 cylinders and then missed all 4 on the next cycle because instead of alternating Lanes  AAAA BBBB AAAA BBBB,  it is more like AAAA ∙ ∙ ∙ ∙ AAAA ∙ ∙ ∙ ∙
I haven't had a chance to test this theory but suggest that if you see it again, shut down one of the lanes and see if it instantly smooths out.
Alternately, start with only one lane energized

Bill Hertzel wrote:

If either lane light is flashing, The ECU is in engine protect mode and battery charging MAY not happen.
The CAN Buss may also disconnect so no engine data will be displayed either.

The battery voltage should never be below 13.0v when the engine is running.
Is is not unusual for the battery voltage to drop into the 10v range during cranking.
Often the computers will Reset or worse...  Partially Reset.
It is not a bad idea to Reset the lanes manually after a start.  It couldn't hurt.  A OFF-ON, Pause, B OFF-ON.

As instructed, be sure to open the throttle half-way before cranking, especially if the engine is not already warmed up.
Yes, it will roar to life and immediately run in the 3000-4000 rpm range.
Keep your hand on the throttle and pull it back as soon as it starts.

I recently had a start where it appeared to be running on 2 cylinders.  Lots of shaking!!!
I suspect that one of the lanes had dropped out but Not Reset and the other lane thought it was still working and what was really happening was it was firing all 4 cylinders and then missed all 4 on the next cycle because instead of alternating Lanes  AAAA BBBB AAAA BBBB,  it is more like AAAA ∙ ∙ ∙ ∙ AAAA ∙ ∙ ∙ ∙
I haven't had a chance to test this theory but suggest that if you see it again, shut down one of the lanes and see if it instantly smooths out.
Alternately, start with only one lane energized

How could one Lane think the other is still working when it really is not? Does it rely on the other lane self-reporting that it has failed or does it try to detect misfires or does it monitor the output of its twin and compare the real output to the expected output?

The Theory here, and it is only a theory, is that while the engine is cranking, the voltage drops far enough for one lane to loose its mind but not far enough for it to reset itself. 
Whatever method the Lane use to cross check each other holds true even though it is not.
It is less than a hundred in one event, but someone ocassionally wins the lottery.
If the theory hold true, then shutting OFF one lane should either kill the engine or it will start running smoothly depending on which lane you selected.
Either result is better than the shaking.
It the engine keeps shaking, then we might not understand what is really going on.

It is worth trying.  There's nothing to lose.