Linear solvers often play paramount roles in large-sale scientific and engineering simulations. In order to maintain the quality of our software's functionalities and performance over various hardwares, our solvers are implemented within the open-source software framework, called Trilinos. However, it is still a challenge to maintain the robust implementation of the parallel solvers, and we often spend significant amount of time maintaining our solvers on each new hardware. Furthermore, in order to effectively utilize the current large-scale computers, many techniques have been proposed (communication avoiding/hiding, asynchronous communication, mixed-precision, random sketching, etc.), which often pose further challenges to maintain the reliability of our solvers. In this talk, we discuss some of these techniques integrated into our solvers, and associated challenges. The hope is to see if there are any potential collaboration in order to enhance the reliabilities of our solver implementations on the current and emerging computers.