About three years ago--New Year's day--transmitter failed (2.5H). Short in the IPA. At that time it was a 2 hour drive for me to the transmitter site. To get the transmitter back on, I connected the MX 15 (which was running at around 16 watts) directly into the final cavity. (In both the "H" and "K" the IPA is connected to the final by a short jumper into a BNC on the side of the PA cavity). Had to turn the exciter DOWN to get back to our (then) licensed TPO of 1100 watts.
So the short answer is yes, you can do this, and it works just fine. As I mention above, 25~30 watts seems to be what is needed to drive the final to full power. You will need to have an exciter that does not mind some reflected power when the transmitter (the PA) is cold, since the input impedance changes as the tube warms up. Otherwise, you may have problems getting back from a power failure. (I notice my Armstrong exciter,--on another transmitter, "fades up" from a power interruption; doesn't seem to bother the tube IPA in that box, a 10K, as it matches the time delay on restart).
To be safe, and to protect the exciter, you need to make a quarter wave shorted stub (piece of RG-8 trimmed for best VSWR, then soldered, works fine) which will protect the exciter in the odd case of a plate to grid short in the final. Stick a PL on it, and "T" it into the line from the exciter to the final. Then just tune the final grid for best match using the exciter's vswr meter.
We replaced our 2.5H3 ("3" for the original TE-3 exciter that it sold with) with a Nautel V-2 recently, but for the last couple of years we've used a 50 watt exciter driving the final 5CX1500B to get our (revised) TPO of 2450 watts. Originally a Continental 802A, then a BE FX-50--the FX-50 is now used to drive the Nautel.
The Nautel appears to need some more babying than the Harris--doesn't like hot weather at our un-air-conditioned site. But that's another thread.