Recall that, for node voltage analysis, a floating voltage source (a voltage source that does not connect to the GND node) poses a problem since you cannot write an equation relating the current through to the voltage across.
What you must do then is enclose the floating voltage source in a supernode, which reduces the number of KCL equations by one, and add the equation relating the voltage difference between the nodes the voltage source is connected to.
Now, the dual of node voltage analysis is mesh current analysis and here we have the dual problem when we have a current source common to two meshes - we can't write an equation relating the current through to the voltage across a current source.
What must be done then is to form a supermesh which reduces the number of KVL equations by one and add the equation relating the difference of the mesh currents to the common current source.
So, write KVL counter-clockwise around the supermesh consisting of the two voltage sources and the two resistors
$$V_1 = I_aR_1 + V_2 + (I_b - I_c)R_2$$
You have, by inspection (no KVL required for this mesh - this is dual to no KCL required for the node connected to a non-floating voltage source)
$$I_c = -1.25A $$
You need one more equation which is the equation relating to difference of the two mesh currents with the common current source.
$$3A = I_a - I_b $$
Now, you have 3 independent equations and 3 unknowns.
You are confused about what the concept of infinity means. Infinity isn't a number that can ever actually measure a quantity of something, like resistance, because it's not a real number. As Wikipedia aptly puts it:
In mathematics, "infinity" is often treated as if it were a number (i.e., it counts or measures things: "an infinite number of terms") but it is not the same sort of number as the real numbers.
When we talk about an "infinite" resistance, what we are really considering is this: as the resistor gets arbitrarily large, what does something (current, voltage, etc) approach?
For example, we can say that as the resistance gets arbitrarily large, current gets arbitrarily small. That is, it approaches zero:
$$ \lim_{R\to\infty} \frac{15\mathrm V}{R} = 0\mathrm{A} $$
That's not the same as saying the current is zero. We can't ever increase R all the way to infinity, so we can't ever decrease current to zero. We can just get arbitrarily close. That means you can't now do this:
$$ \require{cancel} \cancel{0\mathrm A \cdot \infty \Omega = ?}$$
This is a bit of a mathematical contradiction by most definitions of infinity, anyhow. Most numbers, when multiplied by an arbitrarily large number, approach infinity. But, anything multiplied by zero is zero. So when you multiply zero by an arbitrarily large number, what do you get? I haven't a clue. Read more about it on Mathematics.SE: Why is Infinity multiplied by Zero not an easy Zero answer?
You could ask, as the current becomes arbitrarily small, what does the resistance approach?
$$ \lim_{I\searrow 0} \frac{15\mathrm V}{I} = \infty \Omega $$
However, if you look closely, you will notice that if \$I = 0\$, then you are dividing by zero, which is your hint you are approaching something that can't happen. This is why we must ask this question as a one sided limit.
Leaving the realm of mathematics, and returning to the realm of electrical engineering, what do you really get if you remove the resistor from that circuit, and leave it open? What you have now is more like this circuit:
simulate this circuit – Schematic created using CircuitLab
C1 represents the (extremely small) capacitance between the two wires that aren't connected. Really, it was there all along but wasn't significant until the resistance went away. See Why aren't wires capacitors? (answer: they are) and everything has some capacitance to everything else.
Best Answer
We haven't yet covered the general case where R1 and R2 are different. You can use nodal analysis or a Thevenin equivalent, but you can also take a more mathematical approach and use plain KVL. Each path around the circuit gives a different equation. Let's say \$R_1 = 200\ \Omega\$ and \$R_2 = 500\ \Omega\$. We'll define \$I_1\$ as the current through \$R_1\$ and \$I_2\$ as the current through \$R_2\$. There are two possible paths:
$$5\ \mathrm V - 4700\ \Omega \cdot (I_1 + I_2) - 0.6\ \mathrm V - 200\ \Omega \cdot I_1 = 0$$
$$5\ \mathrm V - 4700\ \Omega \cdot (I_1 + I_2) - 0.6\ \mathrm V - 500\ \Omega \cdot I_2 = 0$$
That's two equations in two unknowns, which means we can find a solution. We can solve for the currents by hand or using a convenient tool such as Wolfram Alpha. The solutions are:
$$I_1 \approx 650\ \mathrm{\mu A}$$ $$I_2 \approx 260\ \mathrm{\mu A}$$
Now consider the case where the two resistors are equal \$(R_1 = R_2 = 270\ \Omega)\$.
$$5\ \mathrm V - 4700\ \Omega \cdot (I_1 + I_2) - 0.6\ \mathrm V - 270\ \Omega \cdot I_1 = 0$$ $$5\ \mathrm V - 4700\ \Omega \cdot (I_1 + I_2) - 0.6\ \mathrm V - 270\ \Omega \cdot I_2 = 0$$
We can solve these as simultaneous equations just like before. But notice how similar the equations are! Nothing changes except the name of the variable. But the name of the variable is meaningless -- we can number the currents however we want. Since we can swap the two equations (and thus the two branches) without being able to tell the difference, \$I_1\$ must equal \$I_2\$. We say that the two branches are symmetrical. Symmetry is a powerful concept in mathematics and the sciences. It allows us to simplify our analysis by eliminating variables. If you've studied Gauss's Law in physics, you've already seen symmetry in action.
All this is a fancy mathematical way of saying that when you have two identical branches in parallel, their currents must be the same. That lets us write a single equation:
$$5\ \mathrm V - 4700\ \Omega \cdot (I_{12} + I_{12}) - 0.6\ \mathrm V - 270\ \Omega \cdot I_{12} = 0$$
$$4.4\ \mathrm V - (2 \cdot 4700\ \Omega + 270\ \Omega)I_{12} = 0$$
$$I_{12} \approx 455\ \mathrm{\mu A}$$
where \$I_{12}\$ is the current through either one of the branches.