There are two types of Maintenance Panel codes: progress codes and error codes.  Progress codes are placed in the Maintenance Panel at various stages of initialization.  Error codes are traps which freeze or blink the error number in the maintenance panel.  All errors except the 9000-range errors are fatal.
All MP codes less than 9000 are generated by non-Lisp programs implemented directly in microcode or in Mesa, a high-level programming language in use at Xerox.  MP codes from 9000 to 9999 are generated by an error in the Lisp system code.  Only those MP codes likely to arise during operation pf a Lisp workstation are identified.
0000?0199	Pre-boot diagnostics
0200?0499	Boot diagnostics
0500?0699	IOP codes
0700?0899	Disk diagnostics (written in microcode)
0900?0999	Pilot OS 
1000?6999	Hardware diagnostics
7000?8887	STAR
8888?8888	MP lamp test
9000?9999	Lisp?

0096	Insufficient real memory (<1MByte) for Lisp
0149	Usually right after power-on.  Disk not ready.  Safe and effective to 0-boot from this state.
0200	Normal booting phase 2
0201	CP error in reading from boot device (to recover, try hitting B-RESET once or twice)
0202	Null Mesa germ installed in physical volume
0203	Broken rigid disk boot chain (possibly intermittent)
0204	Illegal IOP port command
0205	CP Trap (CS parity or double-bit memory error)
0206	Null diagnostic microcode in physical volume
0207	Null Pilot/Mesa emulator microcode in physical volume
0208	Null Mesa germ installed in physical volume
0217	Inconsistent Virtual Memory.  Requires reinstallation or try another partition.
0500	StartDomino	Domino has started
0501	InitReadTOD	Domino starting to read the TOD clock
0502	InitReadTODdone	Reading of TOD clock completed (next MP number from Lisp)
0505	CSParity	CS parity error detected
0506	BurdockCPDisabled	Burdock attempted to use EtherKludge
0507	CPBurdockDisabled	CP attempted to use EtherKludge
0508	IOPBreak	An IOP break with no IOP kernel
0509	IllegalIOPIntr	Illegal IOP interrupt
0510	BadMapEntry	Incorrect vm Map entry in IOP access.
0511	NoCPDmaComplete	CP Dma operation failed to complete
0512	NoCPDmaChannel	CP Dma channel not specified
0513	ReadCPPortDead	CP not responding to Read CPPort
0514	WriteCPPortDead	CP not responding to Write CPPort
0520	StackOverflow	A task's stack has overflowed
0565	InvToneCmd	Invalid keyboard tone generator command
0570	InvProcCmd	Invalid command value in Processor CSB
0571	UnImplCmd	Unimplemented command in Processor CSB
0572	SetTODError	The Time-Of-Day could not be set 
0576	LSEPCtlOVR	LSEP Control CSB overrun
0580	NoValidCommand	Invalid floppy IOCB command
0581	UnImplFloppyCmd	Unimplemented floppy IOCB command
0582	InvalidEscapeCmd	Invalid Escape floppy command
0583	CommandTrack	Floppy track register is not correct
0584	TrackToBig	Floppy track number is too large
0585	BadDmaChannel	Couldn't program Floppy Dma
0586	NoDmaEndCount1	External Dma End Count not set
0587	NoDmaEndCount2	Internal Dma End Count not set
0915	Pilot breakpoint	If 0915 occurs during the installation of the software, the problem might be a bad page on the rigid disk, the sys-tem cannot read the floppy disk, or a hardware failure. Try running ALAG.  If ALAG passes and Media Scan shows no new bad pages, the floppy disk (software installation) is suspect.
0937	Trying to find out	Will hang in this state if 
	the time and date	no time server is res-ponding, and the time has not been set on the machine since power-up. Time can be set from within Interlisp (with SETTIME), by 2-booting the Installation Utility floppy, or by 5-booting the EI Fixed Disk Diagnostics floppy).
0981	Trying to discover	Will hang in this state if 
	Ethernet Pup host	non-Lisp  code   tries   to 
	number	perform Pup operations and no Pup ID Server responds or processor ID is not registered.  See network administrator.
Most of these errors are indicative of some serious problem, probably hardware, and usually fatal (but try ???D if you can't TeleRaid).  The main exception is 9004 see description of code 9304.
9001	CSParErr	Control store parity error
9002	StackErr	Hardware stack overflow
9003	IBEmptyErr	Instruction fetch unit empty error
9004	VirtAddrErr	Attempt to reference virtual address >22 bits
9005	EmuMemErr	Double bit memory error or non-existent memory
9013	NegPcError	Inconsistent PC at FnCall
9014	applyUfn	Arg to apply not integer
9016	notFreeTrap	Stack allocation error
9024	Page fault in the page fault handler.
9048	ReFOvr page fault under page fault
9049	Ghost context switch
9051	BadUfnTable 
9120	MiscErr	Opcode no such register
9121	MiscErr	Opcode bad 2nd byte
9127	PcNegError	Inconsistent PC at Punt
9136	CycleMask	Bad caller
9129	M1Loc	Microcode error
9130	M2Loc	Microcode error
9131	M3Loc	Microcode error
These codes generally indicate an error state in Lisp system code that cannot be handled in the break package.  Most are "should never happen" cases that indicate a serious error; but some (in particular, 9305 and 9318) may be much less serious.  If possible, use TeleRaid to find out more information (press the Undo key to enter the TeleRaid server (cursor changes into "TeleRaid"), and run the TeleRaid user from another machine).  Even if you can't TeleRaid from another machine, several of these codes you can convert into a Lisp break if the state of the system is reasonably consistent and the error occurred under user code (rather than, say, the garbage collector): type ???B to the TeleRaid server.
Summary of TeleRaid server commands:
???B	Attempt to enter Break. ? ?If error is in a special system context, will change cursor to "CANT", indicating refusal to enter break. 
???D	Perform Hard Reset-clear stack, flush all non-restartable processes.
???N	Continue from error.  This is usually not possible, except for code 9318, or when you got a 9915 error by typing ???C while the Raid interrupt was enabled.
???P	Display Pup host number (in decimal) in maintenance panel.
The 9300 Series Codes are:
9302	Invalid Vmem: attempt to boot an image that is not a valid Lisp sysout, or which is inconsistent from having some, but not all, of its dirty pages written. Can happen if you boot instead of calling LOGOUT.  Usually caught sooner as code 0217.
9303	No place for IOCB page at startup; should never happen.
9304	Map out of bounds: attempt to use a pointer larger than the virtual address space of the machine.  Usually means garbage was fetched from some-where that should have contained a pointer.  This usually appears as code 9004 instead.
9305	Invalid address: attempt to use a pointer that does not refer to an existing (allocated) part of virtual memory.  Usually means garbage was fetched from somewhere that should have contained a pointer.  This error can often be converted to a break with the ???B TeleRaid command.
9306	Invalid virtual page.  Usually caught sooner, as a 9004. 
9307	Unavailable page on real page chain: inconsistent state in page fault handler.?
9308	Loop in ?SELECTREALPAGE: incon-sistent state in page fault handler.
9309	Attempt to allocate already existing page (from call to ?NEWPAGE).
9310	?DONEWPAGE failed to allocate new map page.
9311	Locked page occupies a file page needed to lock another: bad state in virtual memory system.
9312	Arg to CLOCK0 not an integer box.
9313	Fault on resident page: processor took a page fault for a page that appears to be resident.
9314	PageFault on stack: shouldn't happen, as stack is resident.
9318	Error in uninterruptable system code: an error that ordinarily would enter a break (e.g., a type test failure), but in a piece of code that should not be user-interruptable.  This is generally a sign that some datum used by system code has been smashed, but this is not always fatal.  Should you not have a wizard handy to diagnose the error with TeleRaid, you can type ???N after entering the TeleRaid server; Lisp will go ahead and attempt to enter at the break anyway, from which (if it succeeds) you might be able to glean more information about the problem.
9319	Stack full: hard stack overflow. A soft stack overflow (Lisp break "STACK FULL") occurs when the stack is mostly used up; if you proceed beyond that point without resetting you can completely fill the stack and get this MP code. Press STOP to perform a HARDRESET  to  clear  the stack, or run 
	TeleRaid to find out who was guilty of overflowing the stack.
9320	MDS full: the space for allocation of fixed-length objects is completely exhausted. A continuable Lisp break "STORAGE FULL" occurs when MDS is nearly full.
9321	Unknown UFN: attempt to execute an unimplemented opcode. This usually means that the processor is trying to execute random memory, or took a wild jump somewhere. Often a microcode bug.
9322	Atoms full: the limit on number of litatoms (2?15?) has been reached.
9323	Pnames full: there is no more space for storing the pnames of litatoms. This is unlikely to occur unless you create an extraordinary number of litatoms with very long pnames.
9324	Stack frame use count overflow: the program has attempted to create more than 200 references to the same stack frame.
The only one that is likely to ever occur is 9915, call to RAID.  Note that if you have the Raid interrupt enabled (by default on ???C), you will get a 9915 error by typing that interrupt character.
9905	NOOPSUBR
9906	?BACKGROUNDSUBR
9907	?CHECKBCPLPASSWORD
9908	DISKPARTITION
9909	DSPBOUT
9910	?DSPRATE
9911	?GATHERSTATS
9912	?GETPACKETBUFFER 
9913	?LISPFINISH
9914	?MOREVMEMFILE
9915	RAID
9916	?READRAWPBI 
9917	?WRITERAWPBI
9918	SETSCREENCOLOR
9919	SHOWDISPLAY
9920	?PUPLEVEL1STATE
9921	?WRITESTATS
9922	?CONTEXTSWITCH
9923	?COPYSYS0SUBR
9924	?WRITEMAP