In broad terms, you can use one test pulse to feed as many safety inputs as you want. There's no rules against it, and no reduction in safety rating (unless you assign both channels of one dual-channel safety input device to the same test pulse, in which case your diagnostic coverage is reduced because you can no longer detect cross shorts between channels).
With regard to 1734-IB8S modules specifically, it's open slather. You can assign the same test pulse to every safety input, if you choose, with the caveat above about reduced diagnostic coverage. Typically, with these modules, I assign test pulse 0 to all even-numbered inputs, and test pulse 1 to all odd-numbered inputs. This makes testing and faultfinding in the field without drawings straightforward, with the small trade-off that a fault on one test pulse will take out all four of your dual-channel safety devices. This is often not so much of a problem because if you take out one e/stop, 99% of the time the system is dead in the water anyway and taking out three more won't have any additional effects. If it IS a problem, e.g. with zoned guard switches rather than e/stops, you can use T2 and T3 as well, and then a single test pulse fault only takes out two devices, which if arranged carefully can mitigate the effects fairly well.
The 5069-IB8S safety module, on the other hand, has some restrictions. You still get four test pulses and 8 safety inputs, but it will only allow you to assign each test pulse to a maximum of two inputs. So, in that case, I'm forced to engage all four test pulses to utilise all 8 inputs, and again a single test pulse fault will take out two devices.
The 1756-IB16S is even further locked down. It has 8 test pulse outputs and 16 safety inputs, and each test pulse is permanently assigned to two specific safety inputs and cannot be changed. T0 must be used for i0 and i8, T1 must be used for i1 and i9, and so on. So when using this module, you have no choice but to use the approach above (extrapolated from 8 to 16 inputs), and there's no way to avoid the fact that a fault on one test pulse will take out two safety devices. Careful assignment of safety devices to safety inputs is your only mitigation strategy.
Of course, there's also the consideration that not all safety input devices use the test pulses - inherently self-testing devices like light curtains that have OSSD outputs don't use any test pulses, so that's one factor you can leverage to your advantage if you're trying to minimise the ability for any one test pulse fault to take out multiple devices. Using the additional inputs for single-channel inputs like reset buttons or safety contactor feedback loops as Ken suggested above can provide some mitigation, as taking out your reset button doesn't necessarily shut down a machine - but it could prevent you from resetting it while the fault exists, so it's not a complete solution on its own.