I assume $Q$ is state set of $M$. If so, this runtime bound does not make much sense; in particular, a Turing machine with three states can move to the end of the input and check whether $x$ is an even number; it is clearly not enough to wait three steps.
The next question is whether there is a computable bound on the runtime of halting one-pass machines. Unfortunately, there is not: $M$ might be nondeterministic and halt after an arbitrary amount of steps.
So we determinise $M$ while simulating (that is computable) and look for a bound for all paths. Now we are getting somewhere: after having consumed the input, there are two cases. Either $M$ halts or it enters a loop. As it only moves right (into empty tape), such loops can be detected.
Putting things together, we simulate a determinisation of $M$. We run every branch until $x$ is consumed¹ and then for another $|Q|\cdot|\Sigma_T|$ steps, $Q$ the set of states and $\Sigma_T$ the tape alphabet of $M$. At this point $M$ has either stopped or loops (in this branch), which we detect by a pair of current state and tape symbol occuring for the second time. We only have to check finitely many branches up to a computable bound, therefore $L$ is recursive.
- $M$ might loop here already if not moving the head is allowed, but that can be detected, too.