When listening to most contemporary music, in my experience, time signatures tend to be in "common time", or 4/4. But sometimes it feels like the phrase or measure extends beyond the four quarter notes that comprise the time signature. Is it considered bad form, in terms of music theory (as opposed to "artistic style"), to extend a time signature to twice its normal length?
There is nothing wrong if a phrase goes beyond one bar. You would not change the time signature for that. Time signature is about the pulsation (and marginally the pace) of the music.
But if you feel that a natural unit of what you hear is regularly out of bounds or with wrong accentuation according to its place in the bar, it could mean that the composer has not taken care about that (these days, it might be the most common) or that he precisely meant to have some phrases appear off-balance.
I have seen, albeit not frequently, 4/2 but not 8/4 and I think this is what would fit the situation you allude to (but you don't give much information).
A complete answer would take compound time signatures into account (with periodic modification of bar length) but I prefer that someone more knowledgeable treat this subject.
I remember the moment when the lightbulb went off for me: phrases and bar lines are frequently not aligned.
Consider Haydn Hob XVII:2, variation 1, just to grab what I'm currently learning. The first phrase starts with a pickup (hint), and ends at the second quarter of the 3/4 measure. The last quarter starts the next phrase. Certainly most of the phrases in this piece are aligned with measure boundaries, but most are at least two measures. Other Haydn I can think of is made up of smaller phrases than the measures.
One question to ask is where the accents are. If there is really only one strong pulse every 8 eight notes, then, as I understand it, you might notate 8/8.
When you "listen" to music, how do you know the time signature? After all, time signature is just a means of writing down music in the most intuitive and readable way. As @Faza wrote in his comment, you choose it short enough to be readable and long enough to reflect the rhythmic basis.
Many latin rhythms are actually two-bar patterns and most phrases in music are even longer. It would be difficult to read if the bar had to match the length of the phrase (even neglecting that there are shorter and longer phrases in one piece). The shorter the measures are, the easier it is to read.
However, the measure should reflect the rhyhthmc basis of the piece. Therefore there's a difference between a 6/8 bar and two 3/8 bars, as the felt accent on the "4" of the 6/8 is somewhat less than the accent on the "1".
As others have said, division into measures should reflect the beat you intend, not the phrase boundaries. There is nothing wrong with tying notes across measure boundaries, and sometimes harmonic rules even call for it (suspensions).
Bar lines were not always used historically; medieval and reansissance musicians understood the concept but didn't write it out like we do now. Unless you're writing in a style that specifically calls for self-contained measures all the time (I can't think of one but I'm not up on modern styles), you shouldn't worry about phrases that don't line up with your bar lines if the rhythm does.