Questions about InitializeIMU

[SubodhMishraIdeaForgeTech]

I have questions regarding the priors on Accelerometer and Gyroscope biases. I notice that the InitializeIMU function has been invoked thrice.

Starting with:

ORB_SLAM3/src/LocalMapping.cc

Lines 180 to 187 in 4452a3c

	// Initialize IMU here
	if(!mpCurrentKeyFrame->GetMap()->isImuInitialized() && mbInertial)
	{
	if (mbMonocular)
	InitializeIMU(1e2, 1e10, true);
	else
	InitializeIMU(1e2, 1e5, true);
	}

And then VIBA 1

ORB_SLAM3/src/LocalMapping.cc

Lines 207 to 214 in 4452a3c

	cout << "start VIBA 1" << endl;
	mpCurrentKeyFrame->GetMap()->SetIniertialBA1();
	if (mbMonocular)
	InitializeIMU(1.f, 1e5, true);
	else
	InitializeIMU(1.f, 1e5, true);
	cout << "end VIBA 1" << endl;

and VIBA 2

ORB_SLAM3/src/LocalMapping.cc

Lines 219 to 226 in 4452a3c

	cout << "start VIBA 2" << endl;
	mpCurrentKeyFrame->GetMap()->SetIniertialBA2();
	if (mbMonocular)
	InitializeIMU(0.f, 0.f, true);
	else
	InitializeIMU(0.f, 0.f, true);
	cout << "end VIBA 2" << endl;

I see that the priorsG and A have been reduced from:
(1e2, 1e5) at Initial Initialization to (1.f, 1e5) during VIBA1 to finally (0.0, 0.0) VIBA2.

I wanted to understand the reason behind these choice of parameters and the difference in their values when invoked at different places?

Plus, I noticed that these parameters are used for setting the information matrix, in that case, won't setting it to (0, 0) going to make the prior value completely useless?

Does that mean during VIBA2 you are focussing on estimating Rwg, scale, ba, bg by completely disregarding any priors on ba and bg? If so then why?