Requirements of LAG in a Stacked Deployment

[Architecture]

My company recently purchased a pair of MSR2004-24 routers for one of our sites and I'm having some difficulty configuring a few functions. This is my first experience with Comware and my background is primarily in Cisco devices, so the difference in syntax and certain technologies are things I've yet to fully adapt to. I have a few brief questions that I'm hoping can be cleared up rather quickly by someone with more experience deploying HPE gear.

The routers support stacking via IRF, but appear to only support it over a single one of the routed ports. When I try to add additional ports to the IRF port, the CLI rejects the command and provides the following error:

[HPE-irf-port2/2]port g i g2/0/1
Check failed for reason:
IRF FAIL ifindex not support Bind!

Can someone explain this error to me? Fairly extensive Google searching has yielded little helpful information. Any input would be appreciated. This IRF FAIL message appears on any other port other than GbE X/0/0 on either router. Does IRF care whether a port is set operationally to L2 or L3? My impression, based on documentation, is that IRF supersedes L2 & L3 awareness for the ports on which its configured. Are expansion modules required to configure multi-port IRF on the MSR2004-24?

If I stick with a single physical port configured for IRF, I can successfully get the IRF port up and running.

This leads into me to my next issue: I cannot seem to create an LAG across the two routers into a common switch.  The common switch is a Cisco 2960, set to use LACP to negotiate the LAG. The trunking parameters are the same on both the routers and the switch.

When I try to form the connection, I get the following error: 

%Jan  1 01:08:15:229 2011 Dispatch-GW LAGG/6/LAGG_INACTIVE_HARDWAREVALUE: Member port GE2/0/4 of aggregation group BAGG1 changed to the inactive state, because of the port's hardware restirction.

If I place two ports from the same router in the bridge-agg group, I can successfully form the LAG between the HPE routers and the Cisco switch. Can I not form a LAG using ports from both physical routers? IRF has been confirmed up and running with the dis irf sys-view command.

At first I thought this might be an issue with ports being configured on the slave device, but I can successfully generate a LAG between the routers and the switch (validated with the dis link-a sum command) as long as the physicals ports belong to one router or the other; whether they belong to the slave or master router is irrelevant.

I find it less likely that I've run into the limitations of our hardware and more likely that I simply haven't configured the devices correctly. If anyone could provide some assistance with resolving these issues, I'd be eternally greatful. If you would like to see more of my configuration, please let me know and I will happily paste it here.

Thanks.

[Architecture]

So it looks like the MSR2004 series is limited to IRF on the GbE0/0 interface only. I suppose this is some sort of hardware limitation?

[parnassus]

More than a Hardware limitation it could be called an IRF design (pre)requisite specific of MSR2004 (defined here).

[Architecture]

I had seen that config guide yesterday, but apparently I missed the answer to my second question.

Under the Feature compatibility and configuration restrictions header on page 10, the guide states:

Layer 2 link aggregation is not supported across multiple chassis.

Seems an odd choice to me, but I'm certainly underwhelmed to say the least.