how to code the union or the intersection of different "Dom::Interval" MUPAD objetcts?

I got this example of a symbolic matrix with an "ep" variable that is greater than zero:

[                                               3,   2, 5]
[                                               6,   4, 7]
[                                              ep, 3/2, 0]
[                                   (4*ep - 9)/ep,   7, 0]
[ -(ep*(7*ep - (3*(4*ep - 9))/(2*ep)))/(4*ep - 9),   0, 0]
[                                               7,   0, 0]
 

Where I need to solve this inequation system:

             ep>0
            (4*ep - 9)/ep >0
            -(ep*(7*ep - (3*(4*ep - 9))/(2*ep)))/(4*ep - 9) >0

The way im approaching this is doing first:

            sol=solve((4*ep - 9)/ep >0,ep);   

Where:

vpa(sol)=

 Dom::Interval(-Inf, 0.0)
 Dom::Interval(2.25, Inf)

Then I'm doing the following:

 sol=solve(-(ep*(7*ep - (3*(4*ep - 9))/(2*ep)))/(4*ep - 9) >0,ep);   

Which also produces an undesired answer:

 solve([0.0 < -(1.0*ep*(7.0*ep - (0.5*(12.0*ep - 27.0))/ep))/(4.0*ep - 9.0)], [ep])

that should in fact be :

 vpa(sol)=
 
 Dom::Interval(2.25, Inf)

My questions are:

1. Why does this happen?
2. And supposing it worked, how could I code the union and intersection of this Dom::intervals objects?

For example, I need to intersect the solution intervals obtained from every row, considering the union of those rows that have more than one solution interval, something like this:

 [Dom::Interval(2.25, Inf) U Dom::Interval(-Inf, 0.0) ] intersected with [ Dom::Interval(2.25, Inf)]

That should produce the answer:

 Dom::Interval(2.25, Inf)

I'm creating a Routh Hurwitz stability criterion algoritm and I need to work with symbolic variables and intervals for the special case (first element zero). I need to compute the interval of the positive variable "ep" that keeps the system stable.

Thanks in advance.