While the service manual for the 922A is downloading, let me say this much: I'm suspecting something of a simple nature here because when the amp is bypassed there shouldn't be any input circuitry inline. Matter of fact, in a good design it should be a straight wire from input to output in bypass mode.
Okay, looking at the schematic, this is a very straightforward Grounded Grid Twin Triode amp using a pair of the wonderful big old 3-500Zs. Me likes them.
A neat thing about this design is that the exciter power is added to the output power. That pme pf tje reasons why me likes them.
UNPLUG the amp from AC power and tape the plug ends so you won't plug it in inadvertently. This amp has plenty of safety features, protection circuits and an interlock safety switch but never count on the interlock switch. It is there for a secondary level of protection. Needless to say, the voltages inside this thing can kill you, take precautions.
Looking at page 29 of the pdf service manual, the schematic confirms what I stated above, when the amp is in bypass mode one relay, separate poles ID'd as RL-1-1 and RL-1-2, are indeed a straight wire bypassing the amp and connecting the Input and Output S0-239s.
BUT, there is an RF Volts circuit hanging on there that is used to send current to the meter when the switch for the meter is in the right position.
First thing to check would be the relay contacts themselves, they may be stuck, welded in one position or possibly just need a good cleaning with a strip or two of paper. The Normally Closed set of contacts would be the ones effecting the bypass, examine all contact surfaces in that relay carefully. Of course, blanket replacing a suspect relay is one way, I'd try cleaning contacts and checking them with the ohmmeter first, though.
After that I'd check each of the components in that meter massaging circuit just to the right of the relay contacts on the same page of the schematic. Check C35 and C6 for short, also suspect that one of them may have gone out of range. Then check that diode, D4, with the Diode scale of your multimeter, should read about .5 -.6V in one direction, INF in the other if good.
Doubtful that the resistors could affect the SWR like that but never hurts to give them a quick read if the above yields nothing found.
Then check the lines in there from Input SO-239 to relay for loose connection, cold solder joint, cracked insulator in an SO-239, especially the input side, use the ohmmeter to verify good continuity and no shorts.
Attaching the ohmmeter leads to the center of both Input and Output SOs with alligater clip leads, you can probe in there, power off and disconnected (!) of course, using a spudger to lightly disturb connections, relay, etc. while watching the ohmmeter for signs of bad mechanical connections etc. Doing this and tapping on the relay case with something like a pencil will often show bad, dirty or worn relay contacts as the meter will deflect corresponding to the vibrations you make. In a good relay, there should be no change of the ohmmeter due to mechanical disturbances.
This should prove to be something simple like that.
I'm betting on dirty relay contacts or bad solder joint, possibly bad Input SO-239.
If the SO-239s are fastened with screw and nut hardware in lieu of rivets, make sure they are tight as that is sometimes the ground to chassis. http://kb2ljj.serveftp.com/kenwood/TL-922%20Service%20Manual.pdfhttp://members.fortunecity.com/xe1bef/kenwood-manuals.htm
BTW -- the only online free manual I was able to google up quickly is for the 922 and not the 922A, there may be some differences in the circuit but likely not much in the part we're discussing.
Keep your thinking cap on and if you find anything at all out of the ordinary or something you don't understand, don't force things or get frustrated, stop, come back, ask questions, get answers, you are not under a time limit here, okay?
Always pays to take a look yourself before making the decision as to whether to ship something this size and weight to a shop and all.