Does the font get fixed if you load a new font, e.g. using 'setfont' at the console?  If so, we might be clobbering font memory somehow...

I played around a bit with setfont myself, and found that this patch helped a little (it correctly restored the pre-startup font during VT switch, but clobbered any setting I had done during the VT switch):

diff --git a/src/i830_driver.c b/src/i830_driver.c
index b28b4de..44e9b10 100644
--- a/src/i830_driver.c
+++ b/src/i830_driver.c
@@ -2110,7 +2110,7 @@ SaveHWState(ScrnInfoPtr pScrn)
    }

    vgaHWUnlock(hwp);
-   vgaHWSave(pScrn, vgaReg, VGA_SR_FONTS);
+   vgaHWSave(pScrn, vgaReg, VGA_SR_ALL);

    return TRUE;
 }
@@ -2346,7 +2346,7 @@ RestoreHWState(ScrnInfoPtr pScrn)
    OUTREG(PIPEASTAT, INREG(PIPEASTAT) | FIFO_UNDERRUN);
    OUTREG(PIPEBSTAT, INREG(PIPEBSTAT) | FIFO_UNDERRUN);

-   vgaHWRestore(pScrn, vgaReg, VGA_SR_FONTS);
+   vgaHWRestore(pScrn, vgaReg, VGA_SR_ALL);
    vgaHWLock(hwp);

    return TRUE;

ping response, daniel.

I do not have an easy way to reproduce this bug.
If you can not reproduce it, close it if you want and I will reopen when I find a way to reproduce it reliably.
Anyway I also saw the same problem, in the past weeks, with the same driver in a modern Intel Graphics Media Accelerator X3100 on X86-64.

If you end up reproducing it, can you try using 'setfont' and the patch I posted to see if things behave differently?

Thanks,
Jesse

set as notabug for now and wait for bug reporter to reopen if he can reproduce...