The bugs we are discussing here are unrelated to your configuration, router, switches, PLEX, antennas, etc. They arise from some combination of mistakes made in design and programming , a lack of hardware resources to perform the necessary functions, or both.
When users begin to experience problems which you do not observe as a beta tester, what you are witnessing is an incomplete set of testing that only expose the software to a limited set of test conditions, and therefore did not reveal the problem that occurs in other types of circumstances. A simple example would be that a start of new recording which simultaneously occurred when an EPG update was in progress may cause starting recording to fail, a missed recording.
The bugs that are being reported here on this forum since the firmware update have suddenly changed the way the Tablo behaves, either because of mistakes were made in design or coding, or the hardware lacks the necessary resources to perform the tasks at hand . We are not encountering these missed recordings and undeletable stuck phantom icons 100% of the time. To the contrary, they occur occasionally, but they do occur somewhat repeatedly and in a way that can be replicated .
A technical opinion I expressed a couple years ago on this forum , based on my experience in building real-time systems since 1971, was that a likely explanation was a software "race " design issue. In this situation, synchronization and real-time control sometimes works correctly and other times fails, depending upon how busy the hardware is. In a real-time system, either interrupts or semaphores are used to synchronize two or more processes so as to ensure that they each take place in the correct sequence . If one of the hardware devices lacks the ability and its operating system to support interrupts or semaphores (which is sadly the case with the Roku), then two processes that are supposed to take place in a particular sequence may instead not occur correctly.
If both devices are very fast and have no big latency the chances of a problem are very slight . If on the other hand one or both of the devices are having trouble keeping up with the workload, they will occasionally or frequently fail to synchronize properly.
One bug which lasted on the Roku and Tablo for well over a year caused the Tablo during FForward timeline scrubbing to lose synchronization with the Roku unit crashing and then rebooting . Those of us who lived with that problem found it maddening. Most recently, a single step process of deleting recording which were being made live was changed into a two-step process, with the first step to be stopping the live recording, and then the second step, performed manually at a later time, to actually delete the recording. In both the scrubbing and deletion situations, the solution to the "race" condition is to slow down the processes so that the chance of one going to completion before the second one begins is assured. It is not a coincidence now that the timeline scrubbing now is considerably slower than it used to be, nor is it a coincidence that the deletion of live recordings has been now split into a two-step process . Clearly, there are high demand workloads, particularly of the four tuner unit, which cause certain activities to be slowed down possibly causing them in turn to misbehave. Rearranging task priorities and adding code to inject wait times/delays our time-tested software engineering methods to solve these problems which have been applied successfully here.
Thus, simultaneous downloading of the EPG along with multichannel recording as a hypothetical example, may actually indicate a specific stressful situation which is difficult to be completed 100% of the time without error. Process completion timing and processor workload become especially critical when the device or devices do not support interrupt or semaphore signaling.
Since the Roku and its operating system support neither, the software engineers are left with the challenge of choosing task priorities and deliberate process delay / slowdown code to ensure proper sequence of execution.
I am delighted to learn that the reports of problems have been met with aggressive activity by Tablo to resolve these issues. I certainly do look forward to swift resolution.